test(tty): stabilize no-accumulation test — delta vs pre-loop snapshot of unrelated system sleeps (mount helper sleeps race the window)

test(tty): canonical tool hash parity (bash.ts + await-operation.ts) — same contract as hook parity
feat(tty): MiMo harness process-ownership contract — group-killed bash execution + explicit detachment
2026-09-19 07:24:10 +01:00 · 2026-09-19 07:18:07 +01:00 · 2026-09-19 07:16:35 +01:00 · 2026-09-19 06:39:11 +01:00 · 2026-09-19 06:33:48 +01:00 · 2026-09-19 06:12:38 +01:00
484 changed files with 170719 additions and 6254 deletions
@@ -71,7 +71,13 @@ jobs:
              echo "HEAD is a release bump — loop guard, no re-bump"
              exit 0
            fi
-            WARRANTED=$(python3 scripts/version_convergence.py | python3 -c "import json,sys; print(json.load(sys.stdin).get('warranted_bump') or '')")
+            CONVERGENCE=$(python3 scripts/version_convergence.py)
+            WARRANTED=$(echo "$CONVERGENCE" | python3 -c "import json,sys; d=json.load(sys.stdin); print(d.get('warranted_bump') or '')")
+            CONVERGED=$(echo "$CONVERGENCE" | python3 -c "import json,sys; print(json.load(sys.stdin).get('converged', 'true'))")
+            if [ "$CONVERGED" = "True" ]; then
+              echo "Already converged — no bump needed"
+              exit 0
+            fi
            echo "push convergence: warranted_bump=$WARRANTED"
            if [ -z "$WARRANTED" ]; then
              echo "No substantive commits since last tag — no bump needed"
@@ -179,56 +185,45 @@ jobs:
                  print(f"  - {d}")
              sys.exit(1)

-          # Commit each surface DIRECTLY to master (bookkeeping, not reviewable code)
-          msg_prefix = f"chore(release): auto-bump version to {NEW} — earned by #{PR_NUMBER}"
-          for path, (content, sha) in edited.items():
-              status, r = api(f"/contents/{path}", "PUT", {
-                  "message": f"{msg_prefix} ({PR_TITLE[:80]})\n\nPromotion record: {promoting_url}",
-                  "content": base64.b64encode(content.encode()).decode(),
-                  "sha": sha,
-                  "branch": "master",
-              })
-              if status not in (200, 201):
-                  print(f"  commit {path} failed ({status}): {str(r)[:200]}"); sys.exit(1)
-              print(f"  committed {path} to master")
+          # Commit ALL surfaces as ONE commit via ChangeFiles
+          # (POST /contents with a files[] array). The per-file contents PUT
+          # created one commit per version surface — five same-version "bump
+          # commits" per release (the phantom duplicate-bump cascade,
+          # misdiagnosed as queue drain since 0.66.2; root-caused T15 Part F
+          # 2026-09-11). Git-data blob/tree creation is 404 on this Gitea
+          # build; ChangeFiles is the supported single-commit primitive.
+          msg = f"chore(release): auto-bump version to {NEW} — earned by #{PR_NUMBER} ({PR_TITLE[:80]})\n\nPromotion record: {promoting_url}"
+          ops = [{
+              "operation": "update",
+              "path": path,
+              "content": base64.b64encode(content.encode()).decode(),  # ChangeFileOperation.content is ContentBase64
+          } for path, (content, _old_sha) in edited.items()]
+          status, r = api("/contents", "POST", {
+              "branch": "master",
+              "message": msg,
+              "files": ops,
+          })
+          if status not in (200, 201):
+              print(f"change-files failed ({status}): {str(r)[:300]}"); sys.exit(1)
+          commit_sha = (r.get("commit") or {}).get("sha", "?") if isinstance(r, dict) else "?"
+          print(f"  committed {len(ops)} surfaces as ONE commit {str(commit_sha)[:7]}")

          print(f"AUTO_BUMPED=true — version {NEW} live on master; TUI footer will render v{NEW}")
          with open(os.environ["GITHUB_ENV"], "a") as env:
              env.write(f"RELEASE_VERSION={NEW}\n")
              env.write("AUTO_BUMPED=true\n")
          PYBUMP
-      - name: Create release from merged pull request
+      - name: Create evidence-derived release (SCAFFOLD-TRANCHE-009B)
        env:
          GITEA_TOKEN: ${{ secrets.ROBOT_TOKEN }}
-          PR_TITLE: ${{ gitea.event.pull_request.title }}
-          PR_BODY: ${{ gitea.event.pull_request.body }}
-          RELEASE_API_URL: ${{ gitea.server_url }}/api/v1/repos/${{ gitea.repository }}/releases
-          LIST_API_URL: ${{ gitea.server_url }}/api/v1/repos/${{ gitea.repository }}/releases?limit=50
+          RELEASE_SERVER_URL: ${{ gitea.server_url }}
+          RELEASE_REPO: ${{ gitea.repository }}
        run: |
-          # Skip idempotently if a release for this tag already exists (retry/replay safety).
-          if curl --fail --silent "$LIST_API_URL" \
-            -H "Authorization: token $GITEA_TOKEN" \
-            | python3 -c 'import json,sys,os; exists=[r for r in json.load(sys.stdin) if r.get("tag_name")==f"v{os.environ[\"RELEASE_VERSION\"]}"]; sys.exit(0 if exists else 1)'; then
-            echo "Release v$RELEASE_VERSION already exists — skipping creation"
-            exit 0
-          fi
-          python3 - <<'PY' > /tmp/rune-release.json
-          import json, os
-          version = os.environ["RELEASE_VERSION"]
-          body = f"## {os.environ['PR_TITLE']}\n\n{os.environ.get('PR_BODY', '')}"
-          print(json.dumps({
-              "tag_name": f"v{version}",
-              "target_commitish": "master",
-              "name": f"Rune v{version}",
-              "body": body,
-              "draft": False,
-              "prerelease": False,
-          }))
-          PY
-          curl --fail-with-body --silent --show-error \
-            -X POST \
-            -H "Authorization: token $GITEA_TOKEN" \
-            -H "Content-Type: application/json" \
-            --data-binary @/tmp/rune-release.json \
-            "$RELEASE_API_URL"
-          rm -f /tmp/rune-release.json
+          # ONE note-generation authority: the tested release_notes module.
+          # The old PR-title/body paste path (which produced empty '##'
+          # bodies on direct-push releases) is REMOVED — governed evidence
+          # from the scaffold store is the source, not PR prose.
+          python3 scripts/release_history_cli.py init \
+            --version "$RELEASE_VERSION" \
+            --server "$RELEASE_SERVER_URL" --repo "$RELEASE_REPO" \
+            --store store/scaffold/current.json
@@ -286,12 +286,61 @@ jobs:
        run: |
          NEW="$(echo "${{ gitea.event.pull_request.head.ref }}" | sed 's|chore/auto-bump-||')"
          python3 scripts/check_version_surfaces.py check --root . --version "$NEW"
-      - name: Publish protected-branch gate (full pipeline path)
-        if: ${{ !startsWith(gitea.event.pull_request.head.ref, 'chore/auto-bump') }}
+      - name: Publish protected-branch gate (full pipeline path — all needs green)
+        # RELEASE-SAFETY FIX (2026-09-13): the previous version of this step
+        # read /tmp/need_* files that no step ever writes and echoed success
+        # UNCONDITIONALLY — required-checks published green while type-check
+        # and unit-tests were red, and deploy promoted the red commits
+        # (observed live: 24413e4, fd6e9b4, de30cd9 promoted 0.66.28/29 with
+        # failing type-check). The verdict now comes from the needs context
+        # itself: this step runs ONLY when every required job succeeded.
+        if: >-
+          ${{ !startsWith(gitea.event.pull_request.head.ref, 'chore/auto-bump')
+              && needs.lint.result == 'success'
+              && needs['unit-tests'].result == 'success'
+              && needs.canary.result == 'success'
+              && needs['best-practices'].result == 'success'
+              && needs.documentation.result == 'success'
+              && needs['type-check'].result == 'success'
+              && needs['gitea-offline'].result == 'success' }}
+        run: echo "All required CI checks passed"
+      - name: Refuse promotion when any required check failed (fail-closed)
+        # Companion to the step above: if anything is not exactly success
+        # (failure, cancelled, skipped on a non-bump path), the required-checks
+        # context reports FAILURE — deploy_qualify then refuses promotion.
+        # Fail-closed: not verified → not deployable.
+        if: >-
+          ${{ !startsWith(gitea.event.pull_request.head.ref, 'chore/auto-bump')
+              && (needs.lint.result != 'success'
+                  || needs['unit-tests'].result != 'success'
+                  || needs.canary.result != 'success'
+                  || needs['best-practices'].result != 'success'
+                  || needs.documentation.result != 'success'
+                  || needs['type-check'].result != 'success'
+                  || needs['gitea-offline'].result != 'success') }}
        run: |
-          # All needs must have succeeded (skipped deps = this is a bump PR, handled above)
-          FAILED=""
-          for job in lint unit-tests canary best-practices documentation type-check gitea-offline; do
-            RESULT=$(cat /tmp/need_$job 2>/dev/null || echo "")
-          done
-          echo "All required CI checks passed"
+          echo "::error::Required CI checks not green — refusing to publish required-checks success"
+          echo "lint=${{ needs.lint.result }} unit-tests=${{ needs['unit-tests'].result }} canary=${{ needs.canary.result }} best-practices=${{ needs['best-practices'].result }} documentation=${{ needs.documentation.result }} type-check=${{ needs['type-check'].result }} gitea-offline=${{ needs['gitea-offline'].result }}"
+          exit 1
+
+  notify-event-bridge:
+    # T16 §16C: post the CI completion WAKE to the local event bridge so
+    # event-driven waiters (deploy.yml → ci_event_wait.py) resume on the
+    # event instead of sleep-polling. ALWAYS runs (success or failure —
+    # a red CI must wake the waiter just as loudly as a green one). The
+    # wake is best-effort and never decides the verdict: the waiter
+    # re-reads the canonical Gitea state via deploy_qualify.py.
+    name: CI Pipeline / notify-event-bridge
+    runs-on: rune-ci-desktop
+    needs: [required-checks]
+    if: always()
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          fetch-depth: 1
+      - name: Wake event bridge (wake only — never completion truth)
+        env:
+          CI_SHA: ${{ gitea.sha }}
+          CI_RESULT: ${{ needs.required-checks.result }}
+          RUNE_EVENT_BRIDGE_URL: ${{ vars.RUNE_EVENT_BRIDGE_URL || 'http://127.0.0.1:8791' }}
+        run: python3 scripts/notify_event_bridge.py
@@ -19,38 +19,74 @@ jobs:
        with:
          fetch-depth: 0

-      - name: Wait for CI qualification on this exact SHA
+      - name: Wait for CI qualification on this exact SHA (event-driven, T16 §16C)
+        # T16 §16C/§16D: the old sleep-30 → query → repeat loop is gone.
+        # The waiter registers a correlation on the W/R event bridge and
+        # yields in select() until ci.yml's notify-event-bridge job posts
+        # the completion WAKE; the verdict is then re-read from canonical
+        # Gitea (deploy_qualify.py). The 900s deadline is a FAILURE
+        # DETECTOR (lost-event recovery), never the scheduler.
        env:
          GITEA_SERVER: ${{ gitea.server_url }}
          GITEA_REPO: ${{ gitea.repository }}
          GITEA_TOKEN: ${{ secrets.ROBOT_TOKEN }}
+          # Shared bridge data root: the deploy step runs as the runner
+          # user while the daemon runs as nick — the rendezvous state
+          # (waits/events/resumes/leases) must be shared. §16C live
+          # finding: defaulting to the invoking user's home made the
+          # controller unable to create its FIFO (PermissionError) and
+          # the deploy failed fast+truthfully on both 0.66.63 pushes.
+          RUNE_WR_ROOT: /mnt/workdrive/wr-bridge
+          RUNE_WR_CODE: /home/nick/.local/share/mimo-wait-resume
        run: |
          DEPLOY_SHA=$(git rev-parse HEAD)
          export DEPLOY_SHA
-          echo "Qualifying $DEPLOY_SHA"
-          DEADLINE=$((SECONDS + 900))
-          while [ $SECONDS -lt $DEADLINE ]; do
-            STATUS=$(python3 scripts/deploy_qualify.py)
-            echo "required-checks: $STATUS"
-            if [ "$STATUS" = "success" ]; then exit 0; fi
-            if [ "$STATUS" = "failure" ]; then
-              echo "::error::CI failed on $DEPLOY_SHA — deployment refused"
-              exit 1
-            fi
-            sleep 30
-          done
-          echo "::error::CI qualification timeout on $DEPLOY_SHA — deployment refused"
-          exit 1
+          echo "Qualifying $DEPLOY_SHA (event-driven)"
+          # READINESS BEFORE ACTION (§16U): fail with a clear message if
+          # the bridge rendezvous is not writable in THIS user's context,
+          # instead of a stack trace at FIFO creation time.
+          if ! touch /mnt/workdrive/wr-bridge/resumes/.write-probe 2>/dev/null; then
+            echo "::error::RUNE_WR_ROOT not writable by $(id -un) — bridge data root misconfigured"
+            exit 1
+          fi
+          rm -f /mnt/workdrive/wr-bridge/resumes/.write-probe
+          python3 scripts/ci_event_wait.py --sha "$DEPLOY_SHA" --deadline 900

-      - name: Stage, verify, promote
+      - name: Stage, verify, promote (monotonic)
        env:
          GITEA_TOKEN: ${{ secrets.ROBOT_TOKEN }}
+          GITEA_SERVER: ${{ gitea.server_url }}
+          GITEA_REPO: ${{ gitea.repository }}
        run: |
          python3 scripts/deploy_stage.py

-      - name: Post-deploy runtime verification (the consuming path)
+      - name: Promotion truthfulness check
+        id: promotion_truth
        run: |
          SHA=$(git rev-parse HEAD)
+          CURRENT_TARGET=$(readlink /mnt/workdrive/rune-code/current || true)
+          if [ "$CURRENT_TARGET" = "/mnt/workdrive/rune-code/${SHA:0:12}" ]; then
+            echo "PROMOTED: current == candidate ($SHA)"
+            echo "promoted=true" >> "$GITEA_OUTPUT"
+          else
+            # Not promoted. deploy_stage exit 0 + non-matching current =
+            # NOOP_ALREADY_CURRENT or STALE_SUPERSEDED — truthful non-error
+            # terminal states (a newer build already won). Do NOT claim
+            # promotion; the runtime-verification step below only runs for
+            # real promotions.
+            echo "NOT PROMOTED: current -> $CURRENT_TARGET (superseded/noop — current is authoritative)"
+            echo "promoted=false" >> "$GITEA_OUTPUT"
+          fi
+
+      - name: Post-deploy runtime verification (only for real promotions)
+        if: ${{ success() }}
+        run: |
+          SHA=$(git rev-parse HEAD)
+          CURRENT_TARGET=$(readlink /mnt/workdrive/rune-code/current || true)
+          if [ "$CURRENT_TARGET" != "/mnt/workdrive/rune-code/${SHA:0:12}" ]; then
+            echo "SKIP runtime verification: this job did not promote (superseded/noop)"
+            exit 0
+          fi
          # Exactly what the launcher execs: cd <current>/src, import rune.
          RUNE_HOME=/mnt/workdrive/rune RUNE_MEMORY_DIR=/home/nick/.config/opencode/memory \
            /home/nick/.venvs/memory/bin/python3 -c "
@@ -62,3 +98,28 @@ jobs:
            assert rune.__version__ == info['version']
            print('RUNTIME VERIFIED:', info)
            "
+      - name: Release-history convergence fallback (SCAFFOLD-TRANCHE-0012C)
+        # The PRIMARY convergence mechanism is deploy_stage.py's out-of-lock
+        # publisher (publishes from CURRENT deployment authority, §14B/§14C).
+        # This step is a non-blocking FALLBACK: if the primary publish failed
+        # (e.g. missing Gitea config), this retries. NEVER fails the job —
+        # documentation failure never outweighs deployment validity.
+        if: ${{ success() && steps.promotion_truth.outputs.promoted == 'true' }}
+        continue-on-error: true
+        env:
+          GITEA_TOKEN: ${{ secrets.ROBOT_TOKEN }}
+          RELEASE_SERVER_URL: ${{ gitea.server_url }}
+          RELEASE_REPO: ${{ gitea.repository }}
+        run: |
+          VERSION=$(python3 -c "import sys; sys.path.insert(0,'src'); import rune; print(rune.__version__)" 2>/dev/null || echo "")
+          if [ -z "$VERSION" ]; then
+            echo "SKIP: cannot determine version"
+            exit 0
+          fi
+          python3 scripts/release_history_cli.py refresh \
+            --version "$VERSION" \
+            --server "$RELEASE_SERVER_URL" --repo "$RELEASE_REPO" \
+            --store "$GITHUB_WORKSPACE/store/scaffold/current.json" \
+            --release-sha "$(git rev-parse HEAD)" \
+            --deployed-sha "$(git rev-parse HEAD)" \
+          || echo "::warning::release-history fallback refresh FAILED — in-lock hook may have already converged; retry converges"
@@ -22,9 +22,12 @@ jobs:
    steps:
      - uses: actions/checkout@v4
      - name: Run all canaries
+        # T16 §16Q reconciliation: the glob looped shared/canary_*.py —
+        # real canaries live in src/rune/ (canary_native.MODULES is the
+        # registry authority). The old job silently no-opped.
        run: |
          FAILED=0
-          for canary in shared/canary_*.py; do
+          for canary in src/rune/canary_*.py; do
            if [ -f "$canary" ]; then
              echo "Running $canary..."
              python3 "$canary" || FAILED=$((FAILED + 1))
@@ -11,3 +11,6 @@ build/
 .env
 .venv/
 state/
+# Runtime artifacts (cwd-local event journal + capability index)
+index/
+journal/
@@ -1,6 +1,6 @@
 # Rune

-**Current release: v0.66.4**
+**Current release: v0.86.0**

 **Rune is a persistent personal agent designed to accumulate experience, develop reusable competence, and remain recognisably the same agent even when its underlying foundation model changes.**

@@ -0,0 +1,196 @@
+"""completion_file.py — REAL_ASYNC_BRIDGE: durable completion-file delivery
+with inotify wake. Replaces FIFO+lease for process-wait delivery.
+
+DESIGN:
+  Producer (bridge/daemon/notify):
+    1. Validates operation identity
+    2. Atomically writes a completion file:
+         resumes/<operation_id>.done
+       containing the event/verdict JSON (write to tmp, rename for atomicity)
+
+  Waiter (ci_event_wait / any continuation consumer):
+    1. Checks if the completion file already exists (event-before-waiter)
+    2. If not: registers inotify watch on the resumes/ directory
+    3. Re-checks after watch registration (closes the registration race)
+    4. Blocks in kernel on inotify — zero CPU, zero timer
+    5. On file-created event: reads the file, verifies authority, returns
+
+  The event is a WAKE SIGNAL. Authority still decides terminal truth.
+  The completion file is durable — survives daemon and waiter restarts.
+
+WHY NOT FIFO:
+  FIFO is one-shot (token consumed), requires lease management, and
+  fails under session-busy guards. A completion file is atomic, durable,
+  and works across restarts. inotify wakes the process with zero CPU
+  cost and zero latency between file creation and wake.
+
+WHY NOT POLLING:
+  Polling introduces an arbitrary interval. inotify is a kernel callback
+  that fires exactly when the filesystem changes — the latency tracks
+  the real completion event, not a configured interval.
+"""
+
+from __future__ import annotations
+
+import json
+import os
+import select
+import struct
+import time
+from pathlib import Path
+
+
+def _wr_bridge() -> Path:
+    # WR_ROOT = daemon/infra convention; RUNE_WR_ROOT = repo/scripts
+    # convention. Both honored (WR_ROOT wins) so the daemon, the waiter,
+    # and any isolated test root always resolve the SAME directory.
+    root = os.environ.get("WR_ROOT") or os.environ.get("RUNE_WR_ROOT") or "/mnt/workdrive/wr-bridge"
+    return Path(root)
+
+
+def _resumes_dir() -> Path:
+    return _wr_bridge() / "resumes"
+
+
+# inotify constants (Linux)
+IN_CREATE = 0x100
+IN_CLOSE_WRITE = 0x8
+IN_MOVED_TO = 0x80
+
+
+def completion_path(operation_id: str) -> Path:
+    return _resumes_dir() / f"{operation_id}.done"
+
+
+def write_completion(operation_id: str, event: dict) -> Path:
+    """Atomically write a durable completion file. The rename(2) ensures
+    the waiter never sees a partial file."""
+    _resumes_dir().mkdir(parents=True, exist_ok=True)
+    target = completion_path(operation_id)
+    tmp = target.parent / f".{target.name}.tmp.{os.getpid()}"
+    tmp.write_text(json.dumps(event, sort_keys=True) + "\n")
+    os.rename(str(tmp), str(target))  # atomic on same filesystem
+    os.chmod(str(target), 0o666)  # cross-user bridge
+    return target
+
+
+def read_completion(operation_id: str) -> dict | None:
+    """Read a durable completion file. Returns None if absent."""
+    p = completion_path(operation_id)
+    if not p.is_file():
+        return None
+    try:
+        return json.loads(p.read_text())
+    except (OSError, ValueError):
+        return None
+
+
+def completion_exists(operation_id: str) -> bool:
+    return completion_path(operation_id).is_file()
+
+
+def _inotify_init_and_add_watch(directory: str) -> int:
+    """Create an inotify instance and add a watch on the directory.
+    Returns the inotify fd."""
+    libc = __import__("ctypes").CDLL(None, use_errno=True)
+    fd = libc.inotify_init()
+    if fd < 0:
+        raise OSError("inotify_init failed")
+    # IN_CREATE | IN_MOVED_TO: catch both direct creation and atomic rename
+    mask = IN_CREATE | IN_MOVED_TO
+    wd = libc.inotify_add_watch(fd, directory.encode(), mask)
+    if wd < 0:
+        os.close(fd)
+        raise OSError(f"inotify_add_watch failed for {directory}")
+    return fd
+
+
+def _inotify_read(fd: int, timeout_secs: float | None = None) -> list[str]:
+    """Read pending inotify events. Returns list of filenames that were
+    created/moved-to. Blocks until at least one event or timeout."""
+    poll = select.poll()
+    poll.register(fd, select.POLLIN)
+    events = []
+    buf = os.read(fd, 4096)
+    pos = 0
+    while pos < len(buf):
+        # struct: wd(i32), mask(u32), cookie(u32), name_len(u32)
+        wd, mask, cookie, name_len = struct.unpack_from("iIII", buf, pos)
+        pos += 16
+        name = buf[pos : pos + name_len].rstrip(b"\x00").decode("utf-8", "replace")
+        pos += name_len
+        if mask & (IN_CREATE | IN_MOVED_TO):
+            events.append(name)
+    return events
+
+
+def wait_for_completion(operation_id: str, *, timeout_secs: float | None = None, poll_callback=None) -> dict | None:
+    """Block until a completion file exists for this operation.
+
+    Uses inotify for zero-latency, zero-CPU wake. Falls back to
+    polling_callback (if provided) when inotify is unavailable.
+
+    The completion file is a WAKE SIGNAL. The caller must still verify
+    canonical authority after reading it.
+
+    Returns the event dict from the completion file, or None on timeout.
+    """
+    target = completion_path(operation_id)
+
+    # 1. check-before-watch (event may have arrived before we started)
+    existing = read_completion(operation_id)
+    if existing is not None:
+        return existing
+
+    _resumes_dir().mkdir(parents=True, exist_ok=True)
+
+    try:
+        fd = _inotify_init_and_add_watch(str(_resumes_dir()))
+    except OSError:
+        # inotify unavailable: fall back to polling callback
+        if poll_callback:
+            return poll_callback(operation_id, timeout_secs)
+        return None
+
+    try:
+        # 2. re-check after watch registration (closes the race between
+        #    the initial check and the watch being active)
+        existing = read_completion(operation_id)
+        if existing is not None:
+            return existing
+
+        # 3. block in kernel on inotify — zero CPU, exact latency
+        import select as _select
+
+        poll = _select.poll()
+        poll.register(fd, _select.POLLIN)
+        deadline = (time.monotonic() + timeout_secs) if timeout_secs else None
+
+        while True:
+            remaining = None
+            if deadline is not None:
+                remaining = max(0.1, deadline - time.monotonic())
+                if remaining <= 0.1:
+                    return None  # timeout
+            events = poll.poll(remaining * 1000 if remaining else None)
+            if not events:
+                continue  # poll timeout, re-check
+            names = _inotify_read(fd)
+            if target.name in names:
+                # 4. file appeared — read it
+                existing = read_completion(operation_id)
+                if existing is not None:
+                    return existing
+                # partial/corrupt: keep watching
+    finally:
+        os.close(fd)
+    return None
+
+
+def clear_completion(operation_id: str) -> None:
+    """Remove a consumed completion file (idempotent cleanup)."""
+    p = completion_path(operation_id)
+    try:
+        p.unlink()
+    except OSError:
+        pass
@@ -0,0 +1,799 @@
+#!/usr/bin/env python3
+"""daemon.py — Gitea webhook receiver + reconciliation + resume driver (v0.2).
+
+W/R-v0.2 invariants implemented here:
+  A. SINGLE-FLIGHT PER SESSION — per-session lease (waitstore.acquire_lease,
+     /proc-validated, atomic) + in-process inflight set. A second READY wait
+     for the same session stays queued until the lease releases.
+  B. BUSY SESSION PROTECTION — before spawning, /proc is scanned for a live
+     `mimo run -s <sid>`-shaped process not owned by our lease. Found → skip
+     launch (wait stays READY). No invented MiMo APIs; the lease is the
+     conservative floor against W/R-created concurrency.
+  C. ATOMIC SESSION LEASE — persisted {session_id, wait_id, pid, pgid,
+     pid_starttime, acquired_at, state}; O_EXCL create, steal only on /proc
+     evidence of a dead owner.
+  D. PROCESS TREE OWNERSHIP — workers spawn with start_new_session=True (own
+     process group). Watch timeout ⇒ SIGTERM group → bounded grace → SIGKILL
+     group → verify zero surviving group members → release lease → truthful
+     READY_TO_RETRY (never terminal-on-timeout).
+  E. TIMEOUT IS NOT SUCCESS — rc==0 maps to verdict terminal
+     (COMPLETED/FAILED); rc!=0 or unverifiable ⇒ RESUME_UNVERIFIED
+     (not schedulable, operator-requeueable); timeout ⇒ READY_TO_RETRY.
+  F. STALE RESUMING RECOVERY — on startup, every RESUMING record is
+     inspected: live lease owner ⇒ adopted and monitored; otherwise ⇒
+     RESUME_UNVERIFIED (+ stale lease released). Never auto-relaunched.
+  G. MULTIPLE WAITS, SAME SESSION — schedulable() orders oldest-first; one
+     continuation finishes before the next starts.
+  H. PIDFILE CORRECTNESS — singleton enforced by flock on daemon.pid; the
+     lock dies with the process, so stale pidfiles (incl. SIGKILL) cannot
+     admit a second daemon.
+
+The resume command is configurable (WR_RESUME_CMD, JSON list with
+{session_id}/{msg} placeholders) so the test harness can use a stub; the
+default is the real mechanism: `mimo run -s <sid> -m <model> <msg>`.
+"""
+
+from __future__ import annotations
+import fcntl, json, os, signal, subprocess, sys, threading, time
+from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).parent))
+import waitstore
+import verify
+import completion_file  # REAL_ASYNC_BRIDGE R3: durable completion-file delivery
+
+RESUME_MODEL = "Verboo Code/glm-5.3-flash"
+RECONCILE_SECS = int(os.environ.get("WR_RECONCILE_SECS", "60"))
+RECONCILE_EVERY_TICKS = max(1, int(os.environ.get("WR_RECONCILE_TICKS", "3")))
+WAKE_MIN_AGE_SECS = float(os.environ.get("WR_WAKE_MIN_AGE_SECS", "5"))
+WAKE_RETRIES = int(os.environ.get("WR_WAKE_RETRIES", "3"))
+WAKE_RETRY_SECS = float(os.environ.get("WR_WAKE_RETRY_SECS", "4"))
+DRIVE_LOOP_SECS = int(os.environ.get("WR_DRIVE_SECS", "20"))
+# REAL_ASYNC_BRIDGE §4: a RESUMING wait older than this grace without
+# progress is an orphaned resume flow (thread died between lease acquire
+# and release). Must exceed the longest legitimate in-flight watch
+# (WATCH_SECS); default 3x. Env-overridable for deterministic tests.
+WATCH_SECS = int(os.environ.get("WR_WATCH_SECS", "600"))
+# REAL_ASYNC_BRIDGE §4: a RESUMING wait older than this grace without
+# progress is an orphaned resume flow (thread died between lease acquire
+# and release). Must exceed the longest legitimate in-flight watch
+# (WATCH_SECS); default 3x. Env-overridable for deterministic tests.
+ORPHAN_GRACE_SECS = int(os.environ.get("WR_ORPHAN_GRACE_SECS", str(3 * WATCH_SECS)))
+TERM_GRACE_SECS = int(os.environ.get("WR_TERM_GRACE_SECS", "10"))
+MAX_ATTEMPTS = int(os.environ.get("WR_MAX_ATTEMPTS", "3"))
+_WR_ROOT = Path(os.environ.get("WR_ROOT", "") or Path(__file__).resolve().parent)
+PIDFILE = _WR_ROOT / "daemon.pid"
+LOG = _WR_ROOT / "daemon.log"
+
+DEFAULT_RESUME_CMD = ["mimo", "run", "-s", "{session_id}", "-m", RESUME_MODEL, "{msg}"]
+
+
+def log(msg: str) -> None:
+    with open(LOG, "a") as f:
+        f.write(f"{time.strftime('%H:%M:%S')} {msg}\n")
+
+
+def resume_cmd(session_id: str, msg: str, wait_id: str = "", override: list | None = None) -> list[str]:
+    """T15 §15D: per-wait override wins; placeholders {session_id}/{msg}/
+    {wait_id}. A synthetic controller session (Scaffolding stage driver)
+    is not a MiMo session — the global default cannot resume it, which is
+    exactly why fixed-sleep polling crept back into T12-T14."""
+    if override:
+        return [
+            c.replace("{session_id}", session_id).replace("{msg}", msg).replace("{wait_id}", wait_id) for c in override
+        ]
+    tmpl = os.environ.get("WR_RESUME_CMD")
+    cmd = json.loads(tmpl) if tmpl else DEFAULT_RESUME_CMD
+    return [c.replace("{session_id}", session_id).replace("{msg}", msg).replace("{wait_id}", wait_id) for c in cmd]
+
+
+# ── process-tree evidence & ownership ─────────────────────────────────────
+
+
+def pgid_members(pgid: int) -> list[int]:
+    out = []
+    for d in Path("/proc").iterdir():
+        if not d.name.isdigit():
+            continue
+        try:
+            stat = (d / "stat").read_text()
+            fields = stat.rsplit(")", 1)[1].split()
+            # zombies: killed but not yet reaped. Counting them makes the
+            # survivor check report the direct child we just KILLed before
+            # communicate() reaps it (observed live 2026-09-12).
+            if fields[0] == "Z":
+                continue
+            if int(fields[2]) == pgid:
+                out.append(int(d.name))
+        except (OSError, IndexError, ValueError):
+            continue
+    return out
+
+
+def terminate_tree(pgid: int) -> list[int]:
+    """TERM → bounded grace → KILL. Returns surviving members (must be [])."""
+    try:
+        os.killpg(pgid, signal.SIGTERM)
+    except ProcessLookupError:
+        pass
+    deadline = time.time() + TERM_GRACE_SECS
+    while time.time() < deadline:
+        if not pgid_members(pgid):
+            return []
+        time.sleep(0.2)
+    try:
+        os.killpg(pgid, signal.SIGKILL)
+    except ProcessLookupError:
+        pass
+    for _ in range(50):  # up to 10s for KILL to land
+        if not pgid_members(pgid):
+            return []
+        time.sleep(0.2)
+    return pgid_members(pgid)
+
+
+def stray_worker_pids(session_id: str, exclude_pgids: set[int] = ()) -> list[int]:
+    """Live `mimo run -s <sid>`-shaped processes not owned by our workers."""
+    out = []
+    for d in Path("/proc").iterdir():
+        if not d.name.isdigit():
+            continue
+        try:
+            parts = [p.decode(errors="replace") for p in (d / "cmdline").read_bytes().split(b"\0") if p]
+        except OSError:
+            continue
+        if not parts:
+            continue
+        joined = " ".join(parts)
+        if session_id not in joined or "-s" not in parts:
+            continue
+        if "mimo" not in joined.lower() and "@mimo-ai" not in joined.lower():
+            continue
+        try:
+            if os.getpgid(int(d.name)) in exclude_pgids:
+                continue
+        except OSError:
+            continue
+        out.append(int(d.name))
+    return out
+
+
+# ── verification → transition (unchanged authority model) ─────────────────
+
+
+def _authoritative_verify(rec: dict) -> tuple[str, dict]:
+    """Authoritative verification with a deterministic test seam.
+
+    Production (WR_VERIFY_STUB unset): real Gitea API via verify.py — the
+    ONLY source of success/failure truth. Test (WR_VERIFY_STUB=path to
+    JSON {"<run_ref-or-sha>": verdict, "default": verdict}): stubbed
+    verdicts so invariant tests never touch the network. The stub is a
+    test fixture, not an alternate authority — no production path sets it.
+    """
+    stub_path = os.environ.get("WR_VERIFY_STUB", "")
+    if stub_path:
+        try:
+            spec = json.loads(Path(stub_path).read_text())
+        except Exception:
+            return "pending", {"stub_error": "unreadable stub"}
+        key = str(rec.get("run_ref", "") or rec.get("sha", ""))
+        verdict = spec.get(key, spec.get("default", "pending"))
+        return verdict, {"stub": True, "key": key}
+    import verify
+
+    if rec.get("wait_kind") == "RELEASE_DEPLOY":
+        return verify.verify_release_deploy(rec.get("sha", ""), rec.get("repo_root", ""), rec.get("deployed_link", ""))
+    return verify.verify_run(rec.get("run_ref", ""), rec.get("sha", ""), rec.get("expected", "success"))
+
+
+def try_transition(rec: dict, event: dict | None = None) -> dict:
+    if rec["status"] != "WAITING_EXTERNAL":
+        return rec
+    if rec.get("wait_kind") == "PROCESS_EVENT":
+        # T15 §15D: the matching event IS the authority. The notifier
+        # wrapper observed the real process exit; exit_code decides the
+        # verdict. No Gitea lookup applies. A duplicate event never
+        # reaches here (match applies only to WAITING_EXTERNAL waits).
+        if not event:
+            return rec  # no event → nothing authoritative happened
+        try:
+            exit_code = int(event.get("exit_code"))
+        except (TypeError, ValueError):
+            exit_code = None
+        if exit_code is None:
+            return rec  # malformed event: stay WAITING (15E hostile case)
+        verdict = "success" if exit_code == 0 else "failure"
+        evidence = {"process_event": True, "exit_code": exit_code, "operation_id": event.get("run_ref", "")}
+        log(f"verify {rec['wait_id']}: {verdict} {json.dumps(evidence)[:120]}")
+        rec["status"] = "READY_TO_RESUME"
+        rec["verdict"] = verdict
+        if verdict == "failure":
+            rec["failure_evidence"] = evidence
+        rec["process_event_evidence"] = evidence
+        waitstore.save(rec)
+        return rec
+    verdict, evidence = _authoritative_verify(rec)
+    log(f"verify {rec['wait_id']}: {verdict} {json.dumps(evidence)[:120]}")
+    if verdict == "success":
+        rec["status"] = "READY_TO_RESUME"
+        rec["verdict"] = "success"
+    elif verdict == "failure":
+        rec["status"] = "READY_TO_RESUME"
+        rec["verdict"] = "failure"
+        rec["failure_evidence"] = verify.failed_job_evidence(rec.get("run_ref", ""))
+    else:
+        return rec  # pending/unknown → stay WAITING
+    waitstore.save(rec)
+    return rec
+
+
+def continuation_msg(rec: dict) -> str:
+    c = rec["continuation"]
+    if rec.get("verdict") == "success":
+        return (
+            f"External wait satisfied (wait {rec['wait_id']}).\n"
+            f"Original objective: {c.get('objective', '')}\n"
+            f"Waited for: {c.get('waited_for', '')}\n"
+            f"Authoritative result: workflow success verified via Gitea API.\n"
+            f"SHA/run: {rec.get('sha', '')}/{rec.get('run_ref', '')}\n"
+            f"Verify: {c.get('verify_when_woken', '')}\n"
+            f"Continue from: {c.get('on_success', '')}"
+        )
+    ev = json.dumps(rec.get("failure_evidence", {}))[:400]
+    return (
+        f"External wait finished with FAILURE (wait {rec['wait_id']}).\n"
+        f"Original objective: {c.get('objective', '')}\n"
+        f"Authoritative result: workflow FAILURE verified via Gitea API.\n"
+        f"Failed job evidence: {ev}\n"
+        f"SHA/run: {rec.get('sha', '')}/{rec.get('run_ref', '')}\n"
+        f"Do NOT treat this as success. Continue from: {c.get('on_failure', '')}"
+    )
+
+
+# ── resume driver (single-flight, tree-owned) ─────────────────────────────
+
+
+def resume(rec: dict) -> None:
+    """Run one continuation to completion. Runs on a dedicated worker thread.
+    Preconditions: caller verified no inflight worker for this session."""
+    sid, wid = rec["session_id"], rec["wait_id"]
+    # REAL_ASYNC_BRIDGE §2 invariant: a successful authority-verified
+    # operation completion must NEVER be overwritten or downgraded by a
+    # legacy session/FIFO delivery failure. Terminal operation state
+    # (COMPLETED/FAILED — stamped by the waiter's own authority
+    # verification) is owned by the OPERATION domain; this function owns
+    # only SESSION-DELIVERY. Re-read from disk: the in-memory copy may be
+    # stale relative to the waiter's terminal stamp.
+    fresh = waitstore.load(wid)
+    if fresh and fresh.get("status") in waitstore.TERMINAL:
+        log(
+            f"resume {wid}: skipped — operation already terminal "
+            f"({fresh['status']}); delivery worker never overrides "
+            f"operation truth"
+        )
+        return
+    rec = fresh or rec
+    msg = continuation_msg(rec)
+
+    # B: busy-session guard before touching the lease
+    strays = stray_worker_pids(sid)
+    if strays:
+        rec["status"] = "READY_TO_RESUME"
+        rec["resume_note"] = f"queued: target session busy (stray pids {strays})"
+        waitstore.save(rec)
+        log(f"queued {wid}: session busy {strays}")
+        return
+
+    # pre-register the lease owner as THIS daemon so acquisition is atomic
+    lease = waitstore.acquire_lease(sid, wid, os.getpid(), os.getpid())
+    if lease is None:
+        live = waitstore.read_lease(sid)
+        rec["status"] = "READY_TO_RESUME"
+        rec["resume_note"] = f"queued: lease held by wait {live and live.get('wait_id')}"
+        waitstore.save(rec)
+        log(f"queued {wid}: lease held")
+        return
+
+    rec["status"] = "RESUMING"
+    rec["resume_attempts"] = rec.get("resume_attempts", 0) + 1
+    waitstore.save(rec)
+
+    try:
+        proc = subprocess.Popen(
+            resume_cmd(sid, msg, wid, rec.get("resume_cmd")),
+            stdout=subprocess.PIPE,
+            stderr=subprocess.PIPE,
+            text=True,
+            start_new_session=True,
+        )
+    except Exception as e:
+        waitstore.release_lease(sid, wid)
+        rec["status"] = "READY_TO_RETRY"  # spawn failed pre-delivery → retryable
+        rec["resume_note"] = f"spawn failed: {e}"
+        waitstore.save(rec)
+        log(f"spawn {wid} failed pre-delivery: {e}")
+        return
+
+    pgid = proc.pid  # start_new_session ⇒ pgid == child pid
+    waitstore.update_lease(sid, wait_id=wid, pid=proc.pid, pgid=pgid, pid_starttime=waitstore._pid_starttime(proc.pid))
+    log(f"launched {wid}: pid={proc.pid} pgid={pgid} attempt={rec['resume_attempts']}")
+
+    timed_out = False
+    try:
+        stdout, stderr = proc.communicate(timeout=WATCH_SECS)
+        rc = proc.returncode
+    except subprocess.TimeoutExpired:
+        timed_out = True
+        survivors = terminate_tree(pgid)
+        try:
+            stdout, stderr = proc.communicate(timeout=30)
+        except Exception:
+            stdout, stderr = "", ""
+        rc = proc.returncode
+        if survivors:
+            log(f"ERROR {wid}: group {pgid} survivors after KILL: {survivors}")
+
+    waitstore.release_lease(sid, wid)
+
+    if timed_out:
+        # D+E: timeout is NOT success. Tree terminated; retry truthfully.
+        if rec["resume_attempts"] >= MAX_ATTEMPTS:
+            rec["status"] = "RESUME_UNVERIFIED"
+            rec["resume_note"] = (
+                f"watch timeout x{rec['resume_attempts']}; process group {pgid} terminated; attempts exhausted"
+            )
+        else:
+            rec["status"] = "READY_TO_RETRY"
+            rec["resume_note"] = f"watch timeout; process group {pgid} terminated (TERM→KILL, verified); will retry"
+        log(f"timeout {wid}: tree {pgid} terminated, rc={rc}")
+    elif rc == 0:
+        # COMPLETION TRUTH (constraint 4): rc==0 means the headless worker
+        # EXITED CLEANLY — the strongest terminal evidence the supported CLI
+        # provides. Timeout/cancel can never reach this branch. The evidence
+        # chain (rc, output tail, timing) is recorded on the wait record.
+        terminal = "COMPLETED" if rec.get("verdict") == "success" else "FAILED"
+        rec["status"] = terminal
+        rec["resume_evidence"] = {
+            "kind": "worker_exited_rc0",
+            "rc": 0,
+            "output_tail": (stdout or "")[-400:],
+            "finished_at": waitstore.now(),
+        }
+        log(f"resumed {wid}: rc=0 → {terminal}")
+    else:
+        # Nonzero rc: delivery/completion unproven. Honest, not schedulable.
+        rec["status"] = "RESUME_UNVERIFIED"
+        rec["resume_evidence"] = {
+            "kind": "worker_exited_nonzero",
+            "rc": rc,
+            "output_tail": ((stdout or "") + (stderr or ""))[-400:],
+            "finished_at": waitstore.now(),
+        }
+        rec["resume_note"] = f"resume rc={rc}; delivery unverified"
+        log(f"resumed {wid}: rc={rc} → RESUME_UNVERIFIED")
+    waitstore.save(rec)
+
+
+# ── startup recovery (invariant F) ────────────────────────────────────────
+
+
+def startup_recovery() -> None:
+    for rec in waitstore.all_waits():
+        if rec.get("status") != "RESUMING":
+            continue
+        sid, wid = rec["session_id"], rec["wait_id"]
+        lease = waitstore.read_lease(sid)
+        if lease and lease.get("wait_id") == wid and waitstore._owner_alive(lease):
+            # Adopt the live worker and monitor it to a bounded conclusion.
+            pgid = lease["pgid"]
+            log(f"recover {wid}: adopted live worker pgid={pgid}")
+            deadline = time.time() + WATCH_SECS
+            while time.time() < deadline and pgid_members(pgid):
+                time.sleep(2)
+            survivors = terminate_tree(pgid) if pgid_members(pgid) else []
+            waitstore.release_lease(sid, wid)
+            rec["status"] = "RESUME_UNVERIFIED"
+            rec["resume_note"] = "adopted post-restart; outcome unverifiable" + (
+                f"; survivors killed: {survivors}" if survivors else ""
+            )
+            waitstore.save(rec)
+            log(f"recover {wid}: adopted worker ended → RESUME_UNVERIFIED")
+        else:
+            if lease:
+                waitstore.release_lease(sid, wid)  # stale
+            rec["status"] = "RESUME_UNVERIFIED"
+            rec["resume_note"] = "recovered at daemon startup: no live owner; delivery unverifiable"
+            waitstore.save(rec)
+            log(f"recover {wid}: no live owner → RESUME_UNVERIFIED")
+
+
+# ── reconciler + drive loop (invariant G: deterministic queue) ────────────
+
+_inflight: dict[str, threading.Thread] = {}
+_inflight_lock = threading.Lock()
+
+
+def drive_ready() -> None:
+    for rec in waitstore.schedulable():
+        sid = rec["session_id"]
+        # CONTENTION POLICY (constraint 2): MiMo exposes NO supported
+        # busy/idle session signal (verified 2026-09-12: `mimo session list
+        # --format json` has no status field; no `session status` command;
+        # no status column in the session schema; no documented ACP/serve
+        # endpoint). Therefore automatic injection into interactive sessions
+        # is NOT YET SAFE and stays disabled. Only waits explicitly declared
+        # target_kind="headless" (disposable/owned sessions) auto-dispatch.
+        if rec.get("target_kind") != "headless":
+            if not rec.get("hold_reason"):
+                rec["hold_reason"] = (
+                    "INTERACTIVE_AUTO_RESUME = NOT YET SAFE: "
+                    "no supported MiMo busy/idle signal; "
+                    "queued until operator releases"
+                )
+                waitstore.save(rec)
+            continue
+        if rec.get("ext", {}).get("resume_mode") == "self":
+            # REAL_ASYNC_BRIDGE §3: self-resuming wait (completion-file /
+            # inotify waiter owns its continuation). The daemon has NO
+            # delivery role here — dispatching a delivery worker would
+            # stamp a phantom rc!=0 over the waiter's own truthful
+            # terminal state (the canonical-state contradiction).
+            # NB: keyed on the explicit self marker, NOT on resume_cmd
+            # absence — records without inline resume_cmd may still be
+            # dispatchable via the WR_RESUME_CMD env template.
+            continue
+        with _inflight_lock:
+            t = _inflight.get(sid)
+            if t is not None and t.is_alive():
+                continue  # single-flight: session already being driven
+            # B: never launch into a busy session
+            strays = stray_worker_pids(sid)
+            if strays:
+                log(f"drive: {rec['wait_id']} held, session busy {strays}")
+                continue
+            th = threading.Thread(target=resume, args=(rec,), daemon=True, name=f"resume-{rec['wait_id']}")
+            _inflight[sid] = th
+            # start INSIDE the lock: a just-created thread is not yet
+            # is_alive() until start() runs — starting outside the lock let
+            # the same drive cycle dispatch a second worker for the session
+            # (single-flight race, caught by the T2 invariant test).
+            th.start()
+        log(f"drive: dispatched {rec['wait_id']} (session {sid})")
+
+
+def reconcile_once() -> None:
+    for rec in waitstore.pending():
+        if rec["status"] != "WAITING_EXTERNAL":
+            continue
+        created = rec.get("created_at", "")
+        try:
+            import calendar
+
+            age = time.time() - calendar.timegm(time.strptime(created, "%Y-%m-%dT%H:%M:%SZ"))
+        except ValueError:
+            continue
+        if age > RECONCILE_SECS:
+            try_transition(rec)
+
+
+_tick = 0
+
+
+def orphan_sweep() -> int:
+    """REAL_ASYNC_BRIDGE §4: permanent mid-run orphan-lease recovery.
+
+    FAILURE DETECTION ONLY — never involved in normal completion
+    scheduling. A wait stuck in RESUMING past ORPHAN_GRACE_SECS means the
+    resume flow died between lease acquire and release (worker thread
+    lost, exception mid-watch). Reconcile exactly as startup_recovery
+    would have: release the lease, stamp RESUME_UNVERIFIED truthfully
+    (delivery unverifiable), never auto-relaunch.
+    """
+    import calendar as _cal
+
+    healed = 0
+    for rec in waitstore.all_waits():
+        if rec.get("status") != "RESUMING":
+            continue
+        try:
+            updated = _cal.timegm(time.strptime(rec.get("last_updated", ""), "%Y-%m-%dT%H:%M:%SZ"))
+        except ValueError:
+            continue
+        if time.time() - updated < ORPHAN_GRACE_SECS:
+            continue  # legitimate in-flight watch (up to WATCH_SECS)
+        wid, sid = rec["wait_id"], rec["session_id"]
+        lease = waitstore.read_lease(sid)
+        owner_dead = lease is not None and not waitstore._owner_alive(lease)
+        daemon_owned = lease is not None and lease.get("pid") == os.getpid()
+        if lease is None or owner_dead or daemon_owned:
+            if lease is not None:
+                waitstore.release_lease(sid, wid)
+            if rec.get("resume_attempts", 0) >= MAX_ATTEMPTS:
+                rec["status"] = "RESUME_UNVERIFIED"
+            else:
+                rec["status"] = "READY_TO_RETRY"
+            rec["resume_note"] = (
+                "orphaned mid-run: resume flow died between lease acquire "
+                "and delivery ("
+                + ("lease owner dead" if owner_dead else "resume thread lost" if daemon_owned else "no lease")
+                + "); orphan sweep recovered"
+            )
+            waitstore.save(rec)
+            log(f"orphan-sweep {wid}: {rec['status']} — {rec['resume_note']}")
+            healed += 1
+    return healed
+
+
+def reconciler() -> None:
+    global _tick
+    while True:
+        # v0.3: the local-runner wake is the fast path; periodic
+        # authoritative verification demotes to a slower fallback
+        # (missed-event + daemon-restart recovery), not API polling.
+        _tick += 1
+        if _tick % RECONCILE_EVERY_TICKS == 0:
+            try:
+                reconcile_once()
+            except Exception as e:
+                log(f"reconcile error: {e}")
+            try:
+                orphan_sweep()  # §4 failure detection — bounded, periodic
+            except Exception as e:
+                log(f"orphan-sweep error: {e}")
+        try:
+            drive_ready()
+        except Exception as e:
+            log(f"drive error: {e}")
+        with _inflight_lock:
+            for sid, t in list(_inflight.items()):
+                if not t.is_alive():
+                    del _inflight[sid]
+        time.sleep(DRIVE_LOOP_SECS)
+
+
+# ── webhook receiver (wake signal only) ───────────────────────────────────
+
+
+def wake_sweep(event: dict, eid: str) -> dict:
+    """Sweep WAITING_EXTERNAL waits through authoritative verification.
+
+    Runs on a background thread (the hook's HTTP response returns
+    instantly, minimising the runner's offline window). If any verified
+    wait got 'pending' — the normal case for a job that finished
+    microseconds ago: the post-task hook fires after results are reported
+    but before the runner's final task acknowledgement flips server-side
+    status — retry on a BOUNDED event-triggered schedule (default 4s x3),
+    never a polling loop. The final ack lands milliseconds after the hook
+    exits, so the first retry almost always sees the terminal state.
+    """
+    arrival = time.time()
+    log(f"local wake {eid}: {json.dumps(event)[:140]}")
+    import calendar
+
+    def sweep() -> tuple[int, int, int]:
+        verified = advanced = pending = 0
+        for rec in waitstore.pending():
+            if rec["status"] != "WAITING_EXTERNAL":
+                continue
+            try:
+                age = time.time() - calendar.timegm(time.strptime(rec.get("created_at", ""), "%Y-%m-%dT%H:%M:%SZ"))
+            except ValueError:
+                continue
+            if age < WAKE_MIN_AGE_SECS:
+                continue
+            verified += 1
+            before = rec["status"]
+            try_transition(rec)
+            if rec["status"] != before and rec["status"] == "READY_TO_RESUME":
+                advanced += 1
+            elif rec["status"] == "WAITING_EXTERNAL":
+                pending += 1
+        return verified, advanced, pending
+
+    verified, advanced, pending = sweep()
+    log(
+        f"local wake {eid}: verified={verified} advanced={advanced} "
+        f"pending={pending} sweep_ms={int((time.time() - arrival) * 1000)}"
+    )
+    for attempt in range(1, WAKE_RETRIES + 1):
+        # Retry while the sweep didn't conclusively advance anything:
+        # pending verdicts (ack-window) AND fresh-wait skips (age gate).
+        # Bounded, event-triggered — not a polling loop.
+        if not pending and verified:
+            break
+        time.sleep(WAKE_RETRY_SECS)
+        verified, advanced, pending = sweep()
+        log(
+            f"local wake {eid}: retry{attempt} verified={verified} "
+            f"advanced={advanced} pending={pending} "
+            f"total_ms={int((time.time() - arrival) * 1000)}"
+        )
+    return {"event_id": eid}
+
+
+class Handler(BaseHTTPRequestHandler):
+    def do_POST(self):
+        if self.path == "/process-event":
+            # T15 §15D: process-exit events from builder/controller
+            # notifiers. Minimal dedicated endpoint — the Gitea webhook
+            # path stays untouched. Payload: {operation_id, exit_code,
+            # completed_at} + OPTIONAL provider/event_type/sha (T16
+            # §16C/§16E: the CI adapter registers as provider="gitea",
+            # event_type="ci_completed"; defaults preserve the T15 builder
+            # identity so existing notifiers are unaffected). The event is
+            # a WAKE, never truth — resumed controllers re-read canonical
+            # authority before any verdict.
+            length = int(self.headers.get("Content-Length", 0))
+            body = self.rfile.read(length)
+            try:
+                payload = json.loads(body) if body else {}
+            except json.JSONDecodeError:
+                self.send_response(400)
+                self.end_headers()
+                return
+            event = {
+                "provider": str(payload.get("provider", "builder")),
+                "event_type": str(payload.get("event_type", "builder_exit")),
+                "run_ref": str(payload.get("operation_id", "")) or None,
+                "sha": payload.get("sha"),
+                "exit_code": payload.get("exit_code"),
+                "completed_at": payload.get("completed_at") or "",
+            }
+            event = {k: v for k, v in event.items() if v is not None}
+            eid = waitstore.record_event(event)
+            log(f"process-event {eid}: {json.dumps(event)[:140]}")
+            # REAL_ASYNC_BRIDGE (R3): durable completion-file delivery.
+            # Every event reaching the bridge is written as
+            # resumes/<operation_id>.done so inotify waiters wake on the
+            # kernel event (R4) with no FIFO/lease machinery, and the
+            # event survives daemon AND waiter restarts (R9). Atomic
+            # rename; duplicate events rewrite identical content (R10).
+            _op_id = str(payload.get("operation_id", ""))
+            if _op_id:
+                try:
+                    # Stable identity only: duplicate deliveries of the
+                    # same logical event must produce byte-identical
+                    # completion files (R10 idempotence by construction).
+                    # completed_at is volatile and lives in the durable
+                    # event log; the file is a WAKE SIGNAL, never truth.
+                    _stable = {
+                        k: event.get(k)
+                        for k in ("provider", "event_type", "run_ref", "sha", "exit_code")
+                        if event.get(k) is not None
+                    }
+                    completion_file.write_completion(_op_id, _stable)
+                except Exception as e:  # noqa: BLE001 — delivery best-effort; wait-record transition + watchdog recovery still cover the lost file
+                    log(f"process-event {eid}: completion-file write failed: {e}")
+            matched = 0
+            for rec in waitstore.pending():
+                if rec["status"] == "WAITING_EXTERNAL" and waitstore.match(rec, event):
+                    matched += 1
+                    try_transition(rec, event)
+            # matched READY controller waits are dispatched IMMEDIATELY
+            # (event-driven, no floor). Guard: only never-resumed waits —
+            # a duplicate/late event must not double-dispatch (15E);
+            # single-flight lease + this guard together make resume
+            # idempotent.
+            for rec in waitstore.pending():
+                if rec["status"] == "READY_TO_RESUME" and rec.get("resume_cmd") and rec.get("resume_attempts", 0) == 0:
+                    threading.Thread(target=resume, args=(rec,), daemon=True, name=f"resume-{rec['wait_id']}").start()
+                    matched += 1
+            try:
+                self.send_response(200)
+                self.end_headers()
+                self.wfile.write(json.dumps({"event_id": eid, "matched": matched}).encode())
+            except (BrokenPipeError, ConnectionResetError):
+                pass
+            return
+        if self.path == "/wake":
+            length = int(self.headers.get("Content-Length", 0))
+            body = self.rfile.read(length)
+            try:
+                payload = json.loads(body) if body else {}
+            except json.JSONDecodeError:
+                self.send_response(400)
+                self.end_headers()
+                return
+            event = {
+                "provider": "gitea-local-runner",
+                "task_id": str(payload.get("task_id", "")) or None,
+                "run_id": str(payload.get("run_id", "")) or None,
+                "repository": str(payload.get("repository", "")) or None,
+                "job_result": str(payload.get("job_result", "")) or None,
+            }
+            event = {k: v for k, v in event.items() if v}
+            eid = waitstore.record_event(event)
+            # Respond instantly; sweep on a worker thread. The hook is
+            # fire-and-forget (curl -m 2) — a broken pipe on write is
+            # harmless and must not propagate.
+            try:
+                threading.Thread(
+                    target=wake_sweep, args=(event, eid), daemon=True, name=f"wake-{payload.get('task_id', '?')}"
+                ).start()
+            except Exception as e:
+                log(f"wake spawn error: {e}")
+            try:
+                self.send_response(200)
+                self.end_headers()
+                self.wfile.write(json.dumps({"queued": True, "event_id": eid}).encode())
+            except (BrokenPipeError, ConnectionResetError):
+                pass
+            return
+        if self.path != "/webhook":
+            self.send_response(404)
+            self.end_headers()
+            return
+        length = int(self.headers.get("Content-Length", 0))
+        body = self.rfile.read(length)
+        try:
+            payload = json.loads(body)
+        except json.JSONDecodeError:
+            self.send_response(400)
+            self.end_headers()
+            return
+        action = payload.get("action") or payload.get("event_type") or ""
+        wf = payload.get("workflow_run") or {}
+        event = {
+            "provider": "gitea",
+            "event_type": f"workflow.{action}" if wf else action,
+            "sha": (wf.get("head_sha") or payload.get("sha") or "")[:12] or None,
+            "run_ref": str(wf.get("id") or payload.get("run_number") or "") or None,
+        }
+        event = {k: v for k, v in event.items() if v}
+        eid = waitstore.record_event(event)
+        log(f"webhook event {eid}: {json.dumps(event)[:140]}")
+        matched = 0
+        for rec in waitstore.pending():
+            if rec["status"] == "WAITING_EXTERNAL" and waitstore.match(rec, event):
+                matched += 1
+                try_transition(rec)
+        self.send_response(200)
+        self.end_headers()
+        self.wfile.write(json.dumps({"event_id": eid, "matched": matched}).encode())
+
+    def do_GET(self):
+        if self.path == "/health":
+            self.send_response(200)
+            self.end_headers()
+            self.wfile.write(b'{"ok":true}')
+            return
+        self.send_response(404)
+        self.end_headers()
+
+    def log_message(self, *a):
+        pass
+
+
+def main() -> int:
+    port = int(sys.argv[1]) if len(sys.argv) > 1 else 8791
+
+    # H (constraint 1): flock is the SINGLETON AUTHORITY. daemon.pid's
+    # content is diagnostic metadata only — it never gates startup, never
+    # implies liveness, and is overwritten after the exclusive lock is held.
+    # Stale pid content (incl. SIGKILL debris) is thus harmless by design.
+    PIDFILE.parent.mkdir(parents=True, exist_ok=True)
+    lock_fd = os.open(PIDFILE, os.O_CREAT | os.O_RDWR, 0o644)
+    try:
+        fcntl.flock(lock_fd, fcntl.LOCK_EX | fcntl.LOCK_NB)
+    except OSError:
+        log("refusing start: another daemon holds the pidfile flock")
+        print("daemon already running (pidfile flock held)", file=sys.stderr)
+        return 1
+    os.ftruncate(lock_fd, 0)
+    os.write(lock_fd, f"{os.getpid()}\n".encode())  # diagnostic, lock held for life
+
+    startup_recovery()
+    t = threading.Thread(target=reconciler, daemon=True)
+    t.start()
+    log(f"daemon starting on :{port}")
+    srv = ThreadingHTTPServer(("0.0.0.0", port), Handler)
+    try:
+        srv.serve_forever()
+    except KeyboardInterrupt:
+        pass
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
@@ -0,0 +1,246 @@
+#!/usr/bin/env python3
+"""verify.py — authoritative Gitea verification (webhook ≠ success).
+
+Queries the Gitea API for the TRUE state of the run/job/commit the wait
+references. Returns (verdict, evidence). verdict ∈ {success, failure,
+pending, unknown}. Only this module may move a wait out of
+WAITING_EXTERNAL.
+"""
+
+from __future__ import annotations
+import json, subprocess
+from pathlib import Path
+
+HOST = "https://gitea.thingscouldgetdicey.online"
+REPO = "conductor/rune"
+
+
+def _token() -> str:
+    # Canonical ladder without importing Rune: env → token file → git store
+    import os
+
+    tok = os.environ.get("GITEA_TOKEN", "").strip()
+    if tok:
+        return tok
+    tf = os.path.expanduser("~/.config/opencode/secrets/gitea-token")
+    if os.path.isfile(tf):
+        return open(tf).read().strip()
+    r = subprocess.run(
+        ["git", "credential", "fill"],
+        input="protocol=https\nhost=gitea.thingscouldgetdicey.online\nusername=conductor\n\n",
+        capture_output=True,
+        text=True,
+        timeout=10,
+    )
+    for line in r.stdout.splitlines():
+        if line.startswith("password="):
+            return line[9:]
+    raise RuntimeError("no gitea credential available")
+
+
+def _api(path: str) -> tuple[int, object]:
+    import urllib.request
+
+    req = urllib.request.Request(f"{HOST}/api/v1/repos/{REPO}{path}", headers={"Authorization": f"token {_token()}"})
+    try:
+        with urllib.request.urlopen(req, timeout=20) as r:
+            return r.status, json.loads(r.read())
+    except urllib.error.HTTPError as e:
+        return e.code, {"error": e.read().decode()[:200]}
+
+
+def _jobs_list(run_ref: str):
+    """GET jobs for a run, unwrapping the {'jobs': [...]} envelope."""
+    s, data = _api(f"/actions/runs/{run_ref}/jobs")
+    if s != 200:
+        return s, None
+    if isinstance(data, dict):
+        data = data.get("jobs", [])
+    return s, data if isinstance(data, list) else None
+
+
+def _resolve_run_id(run_ref: str, sha: str) -> str:
+    """Task-list ids are NOT run ids. If run_ref lookup 404s, resolve the
+    real run id by head SHA via /actions/runs (T15: run 6907 vs task 15743)."""
+    if run_ref:
+        s, jobs = _jobs_list(run_ref)
+        if s == 200 and jobs:
+            return run_ref
+    if sha:
+        s, runs = _api(f"/actions/runs?limit=30")
+        items = runs.get("workflow_runs", []) if isinstance(runs, dict) else (runs if isinstance(runs, list) else [])
+        # One SHA fans out to multiple runs (lint + deploy-only). Prefer a
+        # substantive (non-skipped) run; a skipped-only run proves nothing.
+        fallback = ""
+        for r in items:
+            if str(r.get("head_sha", "")).startswith(sha):
+                rid = str(r.get("id"))
+                if r.get("conclusion") != "skipped":
+                    return rid
+                fallback = fallback or rid
+        return fallback
+    return ""
+
+
+def verify_run(run_ref: str, sha: str, expected: str) -> tuple[str, dict]:
+    """run_ref = actions run id (or task id — resolved via SHA lookup).
+
+    Defect fixed 2026-09-12 (surfaced by W/R-v0.3 live proof, run 7110):
+    the jobs query here used raw _api + isinstance(jobs, list), but this
+    Gitea wraps jobs in the {"jobs": [...]} envelope — the check always
+    failed, so EVERY EXACT_WORKFLOW verdict silently fell through to the
+    aggregate commit-status fallback, where unrelated runs on the same
+    SHA (multi-run fan-out: ci.yml + deploy-only) poison the verdict.
+    Use _jobs_list (unwraps the envelope), same as _resolve_run_id."""
+    run_ref = _resolve_run_id(run_ref, sha)
+    if run_ref:
+        s, jobs = _jobs_list(run_ref)
+        if s == 200 and isinstance(jobs, list) and jobs:
+            conclusions = {j.get("name"): j.get("conclusion") for j in jobs}
+            statuses = {j.get("status") for j in jobs}
+            if statuses <= {"completed"}:
+                verdict = "success" if set(conclusions.values()) <= {"success", "skipped"} else "failure"
+                return verdict, {"run_ref": run_ref, "jobs": conclusions}
+            return "pending", {"run_ref": run_ref, "statuses": sorted(statuses)}
+    if sha:
+        s, st = _api(f"/commits/{sha}/status")
+        if s == 200 and isinstance(st, dict):
+            state = st.get("state", "unknown")  # success|error|failure|pending|warning
+            rid = _resolve_run_id("", sha)  # attach run id so failure evidence works
+            if state == "success":
+                return "success", {"sha": sha, "state": state, "run_ref": rid}
+            if state in ("error", "failure"):
+                return "failure", {"sha": sha, "state": state, "run_ref": rid}
+            return "pending", {"sha": sha, "state": state}
+    return "unknown", {"run_ref": run_ref, "sha": sha}
+
+
+def failed_job_evidence(run_ref: str, sha: str = "") -> dict:
+    """For failure continuations: exact failing job + conclusion."""
+    run_ref = _resolve_run_id(run_ref, sha)
+    if not run_ref:
+        return {}
+    s, jobs = _jobs_list(run_ref)
+    if s != 200 or not isinstance(jobs, list):
+        return {}
+    bad = [j for j in jobs if j.get("conclusion") not in ("success", "skipped", None)]
+    return {
+        "failed_jobs": [
+            {"name": j.get("name"), "conclusion": j.get("conclusion"), "url": j.get("url", "")} for j in bad
+        ]
+    }
+
+
+# ── RELEASE_DEPLOY verification (W/R v0.1, overnight handoff Part 2) ──────
+# Aggregate commit status is NOT a valid completion authority for Rune's
+# release/deploy flow: the release successor SHA, not the source SHA, is what
+# converges (observed live: wr-1789179026 stranded on pending while 0.66.15
+# was deployed). The authority chain is:
+#   source SHA → source runs terminal → release successor commit (descendant)
+#   → successor runs success → deployed/current identity converges.
+
+
+def _git(args: list[str], repo_root: str) -> str:
+    import subprocess
+
+    r = subprocess.run(["git"] + args, capture_output=True, text=True, cwd=repo_root or None, timeout=15)
+    if r.returncode != 0:
+        raise RuntimeError(f"git {' '.join(args[:2])}: {r.stderr.strip()[:120]}")
+    return r.stdout
+
+
+def _runs_for_sha(sha: str) -> list[dict]:
+    s, runs = _api("/actions/runs?limit=50")
+    items = runs.get("workflow_runs", []) if isinstance(runs, dict) else []
+    return [r for r in items if str(r.get("head_sha", "")).startswith(sha)]
+
+
+def _sha_verdict_from_runs(runs: list[dict]) -> tuple[str, dict]:
+    """success | failure | pending from a SHA's workflow runs.
+    Terminal verdict ONLY when every run is completed — mixed states
+    (some success, some still in flight) must stay PENDING (observed live:
+    a SHA with 1 completed-success + 1 queued + 1 in-progress was judged
+    success mid-flight; wr-1789182522 2026-09-12)."""
+    if not runs:
+        return "pending", {"note": "no runs found"}
+    non_terminal = [r for r in runs if r.get("status") != "completed"]
+    if non_terminal:
+        return "pending", {"in_flight": len(non_terminal)}
+    conclusions = {r.get("name"): r.get("conclusion") for r in runs}
+    bad = {n: c for n, c in conclusions.items() if c == "failure"}
+    if bad:
+        return "failure", {"failed": bad}
+    return "success", {"runs": conclusions}
+
+
+def verify_release_deploy(source_sha: str, repo_root: str = "", deployed_link: str = "") -> tuple[str, dict]:
+    """RELEASE_DEPLOY authority chain. Returns (verdict, evidence)."""
+    source_sha = (source_sha or "")[:12]
+    ev: dict = {"source_sha": source_sha, "kind": "RELEASE_DEPLOY"}
+
+    # 1. Source SHA's own runs must be terminal and healthy.
+    src_runs = _runs_for_sha(source_sha)
+    src_verdict, src_ev = _sha_verdict_from_runs(src_runs)
+    ev["source"] = src_ev
+    if src_verdict != "success":
+        return src_verdict, ev
+
+    # 2. Release successor: first chore(release) commit descending from source.
+    if not repo_root:
+        return "pending", {**ev, "note": "no repo_root — cannot check lineage"}
+    try:
+        out = _git(
+            [
+                "log",
+                "--format=%H %s",
+                "--reverse",
+                f"{source_sha}..master",
+                "--grep=^chore\\(release\\)",
+                "--extended-regexp",
+            ],
+            repo_root,
+        )
+    except RuntimeError as e:
+        return "pending", {**ev, "note": str(e)}
+    successors = [l for l in out.splitlines() if l.strip()]
+    if not successors:
+        return "pending", {**ev, "note": "no release successor yet"}
+    succ_line = successors[0]
+    succ_sha_full, _, succ_subject = succ_line.partition(" ")
+    succ_sha = succ_sha_full[:12]
+    ev["successor"] = {"sha": succ_sha, "subject": succ_subject[:80]}
+
+    # Ancestry is guaranteed by the range, but verify explicitly (deterministic
+    # lineage over title guessing):
+    try:
+        _git(["merge-base", "--is-ancestor", source_sha, succ_sha_full], repo_root)
+    except RuntimeError:
+        return "failure", {**ev, "note": "successor is not a descendant of source"}
+
+    # 3. Successor runs must be terminal and healthy.
+    succ_runs = _runs_for_sha(succ_sha)
+    succ_verdict, succ_ev = _sha_verdict_from_runs(succ_runs)
+    ev["successor"]["runs"] = succ_ev
+    if succ_verdict != "success":
+        return succ_verdict, ev
+
+    # 4. Deployed identity convergence. A mismatch is FAILURE only when the
+    # successor's runs are ALL completed — if any is still running/queued the
+    # deploy may simply not have propagated yet → PENDING (observed live:
+    # reconciler checked mid-deploy, verdict failure, deployment converged
+    # ~2 min later; wr-1789182522 2026-09-12).
+    if deployed_link:
+        link = Path(deployed_link)
+        if link.is_symlink() or link.is_dir():
+            deployed = str(link.resolve())
+            deployed_id = Path(deployed).name
+            ev["deployed"] = deployed_id
+            if not deployed_id.startswith(succ_sha[:12]):
+                still = [r for r in _runs_for_sha(succ_sha) if r.get("status") != "completed"]
+                if still:
+                    return "pending", {**ev, "note": "deploy propagating (runs in flight)"}
+                return "failure", {**ev, "note": f"deployed {deployed_id} != successor {succ_sha} (all runs terminal)"}
+        else:
+            return "failure", {**ev, "note": f"deployed link missing: {deployed_link}"}
+
+    return "success", ev
@@ -0,0 +1,337 @@
+#!/usr/bin/env python3
+"""waitstore.py — wait/resume state machine v0.2 (developer tooling, NOT Rune).
+
+States:
+  RUNNABLE → WAITING_EXTERNAL → READY_TO_RESUME → RESUMING → COMPLETED | FAILED
+                                       ↑               │
+                                       │               ├─(watch timeout, tree killed)
+                                       │               └─(rc!=0 / unverifiable)
+                                  READY_TO_RETRY ←─────┘
+  RESUME_UNVERIFIED : delivery after crash/timeout cannot be proven.
+                      TERMINAL-ish — never auto-scheduled; explicit
+                      `requeue()` required to make it READY_TO_RESUME.
+  TIMED_OUT         : external event never arrived within the wait budget.
+
+Waits persist as waits/<wait_id>.json (atomic tmp+rename).
+Leases persist as leases/<session_id>.lease.json (atomic tmp+rename,
+validated against /proc so a dead or PID-reused owner is stealable).
+"""
+
+from __future__ import annotations
+import fcntl, json, os, tempfile, time, uuid
+from pathlib import Path
+
+ROOT = Path(os.environ.get("WR_ROOT", "") or Path(__file__).resolve().parent)
+WAITS = ROOT / "waits"
+EVENTS = ROOT / "events"
+LEASES = ROOT / "leases"
+
+SCHEDULABLE = {"READY_TO_RESUME", "READY_TO_RETRY"}
+VALID_STATES = {
+    "RUNNABLE",
+    "WAITING_EXTERNAL",
+    "READY_TO_RESUME",
+    "RESUMING",
+    "COMPLETED",
+    "FAILED",
+    "READY_TO_RETRY",
+    "RESUME_UNVERIFIED",
+    "TIMED_OUT",
+}
+TERMINAL = {"COMPLETED", "FAILED", "RESUME_UNVERIFIED", "TIMED_OUT"}
+
+
+def _atomic_write(path: Path, data: dict) -> None:
+    path.parent.mkdir(parents=True, exist_ok=True)
+    fd, tmp = tempfile.mkstemp(dir=str(path.parent), prefix=".tmp-")
+    try:
+        with os.fdopen(fd, "w") as f:
+            json.dump(data, f, indent=1, sort_keys=True)
+        os.replace(tmp, path)
+        # T17 17C: the bridge root is SHARED across users (deploy waiter
+        # runs as the runner user; daemon as nick). mkstemp creates 0600 —
+        # daemon re-saves locked the runner user out of records it must
+        # read (waitstore.load in await_resume_event; find_matching_event
+        # replay scan silently skipped them). 0644: owner writes, all read.
+        os.chmod(path, 0o644)
+    except BaseException:
+        if os.path.exists(tmp):
+            os.unlink(tmp)
+        raise
+
+
+def now() -> str:
+    return time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime())
+
+
+def new_wait(
+    *,
+    session_id: str,
+    provider: str,
+    event_type: str,
+    repo: str,
+    sha: str = "",
+    run_ref: str = "",
+    expected: str = "success",
+    wait_kind: str = "EXACT_WORKFLOW",
+    target_kind: str = "interactive",
+    repo_root: str = "",
+    deployed_link: str = "",
+    objective: str = "",
+    waited_for: str = "",
+    verify_when_woken: str = "",
+    on_success: str = "",
+    on_failure: str = "",
+    resume_cmd: list | None = None,
+) -> dict:
+    wid = f"wr-{int(time.time())}-{uuid.uuid4().hex[:6]}"
+    rec = {
+        "wait_id": wid,
+        "session_id": session_id,
+        "created_at": now(),
+        "provider": provider,
+        "event_type": event_type,
+        "repo": repo,
+        "sha": sha,
+        "run_ref": run_ref,
+        "expected": expected,
+        # T15 §15D: PROCESS_EVENT — a wait whose authoritative verdict is
+        # the matching process-exit event itself (exit_code carried on the
+        # event), e.g. a builder/controller process completing. The event
+        # IS the authority here (the notifier wrapper observed the real
+        # process exit); no Gitea lookup applies.
+        "wait_kind": wait_kind
+        if wait_kind in ("EXACT_WORKFLOW", "RELEASE_DEPLOY", "PROCESS_EVENT")
+        else "EXACT_WORKFLOW",
+        # CONTENTION POLICY: "interactive" (default, conservative) never
+        # auto-dispatches — MiMo exposes NO supported busy/idle session
+        # signal, so injecting into a possibly-busy interactive session is
+        # NOT YET SAFE. Only explicitly declared headless/disposable
+        # targets ("headless") are auto-resumed.
+        "target_kind": target_kind if target_kind in ("headless", "interactive") else "interactive",
+        "repo_root": repo_root,
+        "deployed_link": deployed_link,
+        # T15 §15D: per-wait resume command override (placeholder template,
+        # same substitution as WR_RESUME_CMD: {session_id}/{msg}/{wait_id}).
+        # A synthetic controller session (e.g. a Scaffolding stage driver)
+        # is NOT a MiMo session — the global mimo-shaped default cannot
+        # resume it, which is exactly why fixed-sleep polling crept back
+        # in T12-T14. Per-wait override makes ANY registered continuation
+        # target resumable through the same bridge.
+        "resume_cmd": list(resume_cmd) if resume_cmd else None,
+        "status": "WAITING_EXTERNAL",
+        "last_verified": "",
+        "resume_attempts": 0,
+        "continuation": {
+            "objective": objective,
+            "waited_for": waited_for,
+            "verify_when_woken": verify_when_woken,
+            "on_success": on_success,
+            "on_failure": on_failure,
+        },
+    }
+    _atomic_write(WAITS / f"{wid}.json", rec)
+    return rec
+
+
+def load(wid: str) -> dict | None:
+    p = WAITS / f"{wid}.json"
+    return json.loads(p.read_text()) if p.is_file() else None
+
+
+def save(rec: dict) -> None:
+    assert rec.get("status") in VALID_STATES, rec.get("status")
+    rec["last_updated"] = now()
+    _atomic_write(WAITS / f"{rec['wait_id']}.json", rec)
+
+
+def all_waits() -> list[dict]:
+    out = []
+    for p in WAITS.glob("*.json"):
+        try:
+            out.append(json.loads(p.read_text()))
+        except json.JSONDecodeError:
+            continue
+    return out
+
+
+def pending() -> list[dict]:
+    return [r for r in all_waits() if r.get("status") in {"WAITING_EXTERNAL"} | SCHEDULABLE]
+
+
+def schedulable() -> list[dict]:
+    """READY waits, oldest first — deterministic queue order."""
+    rs = [r for r in all_waits() if r.get("status") in SCHEDULABLE]
+    rs.sort(key=lambda r: (r.get("created_at", ""), r.get("wait_id", "")))
+    return rs
+
+
+def requeue(wid: str, note: str) -> dict | None:
+    """Operator-explicit requeue of a RESUME_UNVERIFIED/RESUME_FAILED wait."""
+    rec = load(wid)
+    if not rec:
+        return None
+    rec["status"] = "READY_TO_RESUME"
+    rec["resume_note"] = note
+    save(rec)
+    return rec
+
+
+# ── events (wake signals, idempotent) ─────────────────────────────────────
+
+
+def record_event(event: dict) -> str:
+    import hashlib
+
+    body = json.dumps(event, sort_keys=True)
+    eid = hashlib.sha256(body.encode()).hexdigest()[:16]
+    f = EVENTS / f"{eid}.json"
+    if not f.is_file():
+        _atomic_write(f, {"event_id": eid, "received_at": now(), "event": event})
+    return eid
+
+
+def match(rec: dict, event: dict) -> bool:
+    if rec.get("provider") != event.get("provider"):
+        return False
+    if rec.get("event_type") and event.get("event_type") and rec["event_type"] != event["event_type"]:
+        return False
+    if rec.get("sha") and event.get("sha") and rec["sha"] != event["sha"]:
+        return False
+    if rec.get("run_ref") and event.get("run_ref") and rec["run_ref"] != event["run_ref"]:
+        return False
+    return True
+
+
+# ── leases (per-session single-flight, /proc-validated) ───────────────────
+#
+# Invariant 3 (PID reuse): liveness is NEVER "PID exists". The lease records
+# pid + pgid + process starttime (/proc/<pid>/stat field 22); all three must
+# match or the owner is a different process wearing a recycled PID.
+#
+# Invariant 1 (flock as authority): the lease file's flock serializes
+# read-modify-write across processes and dies with its holder, so a crashed
+# daemon can never wedge the lock. The lease JSON + /proc decide who OWNS the
+# session; the flock only prevents racing acquirers from double-stealing.
+
+_lease_fds: dict[str, int] = {}  # session_id → open fd holding the flock
+
+
+def _pid_starttime(pid: int) -> str | None:
+    """Field 22 of /proc/<pid>/stat — distinguishes PID reuse."""
+    try:
+        stat = Path(f"/proc/{pid}/stat").read_text()
+        return stat.rsplit(")", 1)[1].split()[19]
+    except (OSError, IndexError):
+        return None
+
+
+def _owner_alive(lease: dict) -> bool:
+    pid = lease.get("pid")
+    if not pid or not Path(f"/proc/{pid}").is_dir():
+        return False
+    if _pid_starttime(pid) != lease.get("pid_starttime"):
+        return False  # PID reused by an unrelated process
+    try:
+        if os.getpgid(pid) != lease.get("pgid"):
+            return False
+    except OSError:
+        return False
+    return True
+
+
+def lease_path(session_id: str) -> Path:
+    # T16 §16C finding (observed live in daemon.log): session ids like
+    # "controller-<op>-Verboo Code/glm-5.3-flash" contain "/" — the raw
+    # embed made lease creation raise FileNotFoundError, so resume
+    # delivery for those waits could never succeed. Sanitize to a flat
+    # filesystem-safe name; the hash suffix prevents collisions between
+    # ids that sanitize identically.
+    import hashlib as _hashlib
+    import re as _re
+
+    safe = _re.sub(r"[^A-Za-z0-9._-]+", "_", session_id)[:80]
+    digest = _hashlib.sha256(session_id.encode()).hexdigest()[:10]
+    return LEASES / f"{safe}-{digest}.lease.json"
+
+
+def read_lease(session_id: str) -> dict | None:
+    p = lease_path(session_id)
+    if not p.is_file():
+        return None
+    try:
+        return json.loads(p.read_text())
+    except json.JSONDecodeError:
+        return None
+
+
+def _locked_lease_fd(session_id: str) -> int | None:
+    """Open the lease file and take its flock (non-blocking). The fd stays
+    open (lock held) until release_lease closes it."""
+    if session_id in _lease_fds:
+        return _lease_fds[session_id]  # already hold it
+    LEASES.mkdir(parents=True, exist_ok=True)
+    fd = os.open(lease_path(session_id), os.O_CREAT | os.O_RDWR, 0o600)
+    try:
+        fcntl.flock(fd, fcntl.LOCK_EX | fcntl.LOCK_NB)
+    except OSError:
+        os.close(fd)
+        return None  # another process holds the lock — mid-acquire
+    _lease_fds[session_id] = fd
+    return fd
+
+
+def _write_lease_locked(session_id: str, lease: dict) -> None:
+    fd = _lease_fds[session_id]
+    os.ftruncate(fd, 0)
+    os.lseek(fd, 0, os.SEEK_SET)
+    os.write(fd, json.dumps(lease, indent=1, sort_keys=True).encode())
+    os.fsync(fd)
+
+
+def acquire_lease(session_id: str, wait_id: str, pid: int, pgid: int) -> dict | None:
+    """Atomically create or steal the session lease. Returns the lease dict,
+    or None if a live owner holds it (or another process is mid-steal).
+    Staleness is decided by /proc identity evidence, never by age alone."""
+    fd = _locked_lease_fd(session_id)
+    if fd is None:
+        return None
+    existing = read_lease(session_id)
+    if existing and _owner_alive(existing):
+        _lease_fds.pop(session_id, None)
+        os.close(fd)
+        return None  # live worker owns this session
+    lease = {
+        "session_id": session_id,
+        "wait_id": wait_id,
+        "pid": pid,
+        "pgid": pgid,
+        "pid_starttime": _pid_starttime(pid),
+        "acquired_at": now(),
+        "state": "RESUMING",
+    }
+    _write_lease_locked(session_id, lease)
+    return lease
+
+
+def update_lease(session_id: str, **fields) -> None:
+    lease = read_lease(session_id) or {}
+    lease.update(fields)
+    _write_lease_locked(session_id, lease)
+
+
+def release_lease(session_id: str, wait_id: str) -> bool:
+    """Release only if the current lease belongs to this wait (no stealing
+    a successor's lease during a slow shutdown path). Closes the flock fd."""
+    lease = read_lease(session_id)
+    if not (lease and lease.get("wait_id") == wait_id):
+        return False
+    fd = _lease_fds.pop(session_id, None)
+    if fd is not None:
+        try:
+            fcntl.flock(fd, fcntl.LOCK_UN)
+        except OSError:
+            pass
+        os.close(fd)
+    lease_path(session_id).unlink(missing_ok=True)
+    return True
@@ -0,0 +1,56 @@
+import type { Hooks } from "@mimo-ai/plugin"
+
+// event-wait-guard — RETIRED (T21.5 §6/§8, 2026-09-17).
+//
+// HISTORY: this hook was the T21 §21B-3 runtime interception layer that
+// blocked model-generated sleep→poll continuation (T20 regression), and
+// later carried the T21.1 §21.1C cwd guard. Both were DOWNSTREAM patches:
+// they corrected the model's output instead of fixing the generator.
+//
+// ROOT-CAUSE FIX (T21.5 §2-§5): the generator was ci_event_wait's
+// wait loop returning nonterminal control to the model — caused by a
+// single hard-coded master-deploy authority that was blind to branch/PR
+// operations. T21.5 fixed this upstream:
+//   - OPERATION_SPECS: per-operation-type completion authority
+//     (branch_ci/pull_request_checks -> own Actions runs;
+//     master_ci/deploy/release -> deploy qualification);
+//   - SINGLE LOGICAL AWAIT: all nonterminal conditions re-arm inside
+//     ci_event_wait.run(); the model sees only terminal results;
+//   - structured PendingOperation envelopes (op_type, authority,
+//     internal_rearm_count) propagate to the controller.
+//
+// With the generator fixed, runtime interception is retired per the
+// tranche rule FIXED UPSTREAM BEHAVIOUR > DOWNSTREAM GUARD CORRECTION.
+// Regression prevention lives in the test suite:
+//   tests/test_single_logical_await.py  (no model re-arm contract)
+//   tests/test_ci_event_wait.py         (watchdog=recovery, event=wake)
+//   tests/test_wait_guard_corpus.py     (canonical wait-pattern corpus)
+// NOT as a permanent runtime block. The hook file stays installed as a
+// documented no-op so canonical-source/install parity (test_hook_drift,
+// test_hook_parity) continues to prove the installed runtime matches
+// this repository's canonical state.
+
+export const EventWaitGuard: Hooks = {
+  "tool.execute.before": async (_input, _output) => {
+    // RETIRED — no interception. See history above.
+    return
+  },
+}
+
+export const CwdGuard: Hooks = {
+  "tool.execute.before": async (_input, _output) => {
+    // RETIRED (T21.5 §8) — the harness/construction execution boundary
+    // now resolves project context upstream (execution-context binding).
+    // No runtime refusal. See tests/test_auto_execution_context.py.
+    return
+  },
+}
+
+export const CombinedGuards: Hooks = {
+  "tool.execute.before": async (input, output) => {
+    // Kept as the default export shape; both sub-guards are retired.
+    return
+  },
+}
+
+export default CombinedGuards
@@ -0,0 +1,92 @@
+import { tool } from "@mimo-ai/plugin"
+import { execSync } from "child_process"
+
+export default tool({
+    description: "Wait for an external operation (CI, deploy, release) to complete. Call this when you cannot make progress until an external process finishes. You do NOT need to know the sha, op_type, or duration — Rune's controller resolves those from the bound PendingOperation. Just call this when you want to wait.",
+    args: {
+        operation_id: tool.schema.string().optional().describe("Optional: the operation to wait for. If omitted, the controller resolves the current pending operation automatically."),
+        task_id: tool.schema.string().optional().describe("Optional: the active task id (e.g. T4). When provided, resolution is scoped to operations bound to this task — stale global operations are never consulted."),
+        sha: tool.schema.string().optional().describe("OPTIONAL HINT — non-authoritative. The controller uses this only to help resolve which operation you mean."),
+        op_type: tool.schema.string().optional().describe("OPTIONAL HINT — non-authoritative. The controller derives the canonical operation type from repository/Gitea state."),
+        deadline_secs: tool.schema.number().optional().describe("OPTIONAL HINT — non-authoritative. The controller applies governed policy timing. Real user deadlines must be stated in conversation."),
+    },
+    async execute(args) {
+        // INTENT NORMALIZATION — model output is intent; the controller
+        // owns execution semantics. All model args are hints recorded for
+        // observability; canonical truth comes from the bound operation.
+        const parts = []
+        if (args.operation_id) parts.push(`--operation-id ${args.operation_id}`)
+        if (args.task_id) parts.push(`--task-id ${args.task_id}`)
+        if (args.sha) parts.push(`--sha ${args.sha}`)
+        if (args.op_type) parts.push(`--model-op-type ${args.op_type}`)
+        if (args.deadline_secs) parts.push(`--model-deadline ${args.deadline_secs}`)
+        const normCmd = `python3 /mnt/workdrive/issue161-harness/scripts/await_intent_cli.py normalize ${parts.join(" ")}`
+        let norm: any
+        try {
+            const normOut = execSync(normCmd, {
+                cwd: "/mnt/workdrive/issue161-harness",
+                encoding: "utf-8" as const,
+                timeout: 120000,
+                env: { ...process.env, RUNE_WR_ROOT: "/mnt/workdrive/wr-bridge" },
+            })
+            norm = JSON.parse(normOut.trim().split("\n").find((l: string) => l.startsWith("{")) || "{}")
+        } catch (e: any) {
+            return JSON.stringify({ ok: false, stage: "normalization", error: (e.stderr || e.message || "?").slice(0, 300) })
+        }
+        if (norm.normalized === "AMBIGUOUS") {
+            return JSON.stringify({ ok: false, stage: "normalization", ...norm })
+        }
+        if (!norm.normalized) {
+            return JSON.stringify({ ok: false, stage: "normalization", ...norm })
+        }
+        // canonical continuation: stored/derived op_type + POLICY deadline
+        // (the sha for the wait comes from the normalization resolution,
+        // NOT from the model's args)
+        const waitSha = norm.sha_for_wait || args.sha || ""
+        if (!waitSha) {
+            return JSON.stringify({ ok: false, stage: "resolution", error: "no sha resolved for the pending operation", norm })
+        }
+        const cmd = `python3 /mnt/workdrive/issue161-harness/scripts/ci_event_wait.py --sha ${waitSha} --deadline ${norm.deadline_secs} --op-type ${norm.op_type}`
+        try {
+            const result = execSync(cmd, {
+                cwd: "/mnt/workdrive/issue161-harness",
+                timeout: (norm.deadline_secs + 60) * 1000,
+                encoding: "utf-8" as const,
+                env: { ...process.env, DEPLOY_SHA: waitSha, RUNE_WR_ROOT: "/mnt/workdrive/wr-bridge", RUNE_WR_CODE: "/home/nick/.local/share/mimo-wait-resume" },
+            })
+            const jl = result.trim().split("\n").find((l: string) => l.startsWith("{"))
+            const envelope = jl ? JSON.parse(jl) : { ok: false, error: "no output" }
+            return JSON.stringify({
+                ...envelope,
+                normalization: {
+                    model_deadline_ignored: norm.model_deadline_ignored,
+                    model_op_type_ignored: norm.model_op_type_ignored,
+                    model_sha_ignored: !!args.sha && args.sha !== waitSha,
+                    canonical_op_type: norm.op_type,
+                    canonical_authority: norm.authority,
+                    deadline_provenance: norm.deadline_provenance,
+                    matched_by: norm.matched_by || norm.evidence || "auto-resolved",
+                },
+            })
+        } catch (e: any) {
+            const lines = (e.stdout || "").trim().split("\n")
+            const jl = lines.find((l: string) => l.startsWith("{"))
+            if (jl) {
+                const envelope = JSON.parse(jl)
+                return JSON.stringify({
+                    ...envelope,
+                    normalization: {
+                        model_deadline_ignored: norm.model_deadline_ignored,
+                        model_op_type_ignored: norm.model_op_type_ignored,
+                        model_sha_ignored: !!args.sha && args.sha !== waitSha,
+                        canonical_op_type: norm.op_type,
+                        canonical_authority: norm.authority,
+                        deadline_provenance: norm.deadline_provenance,
+                        matched_by: norm.matched_by || norm.evidence || "auto-resolved",
+                    },
+                })
+            }
+            return JSON.stringify({ ok: false, stage: "wait", error: (e.stderr || e.message || "?").slice(0, 300) })
+        }
+    },
+})
@@ -0,0 +1,167 @@
+
+// bash — executes a shell command. PROCESS-OWNERSHIP CONTRACT (v2):
+//
+//   MODE A (tool-owned): the child is launched in its own process group;
+//   stdout/stderr are CAPTURED pipes (never the TUI tty). On timeout or
+//   failure the ENTIRE process group is signalled (SIGTERM, grace,
+//   SIGKILL) so grandchildren cannot survive the invocation, then the
+//   result is finalized. After finalization NO descendant retains a
+//   writer into MiMo's streams.
+//
+//   MODE B (explicitly detached): a command suffixed with " &" that the
+//   user/model intends to outlive the tool (nohup/setsid/disown-style)
+//   is executed via the detached helper: stdin=/dev/null, stdout/stderr
+//   -> durable log file, own session (setsid), no MiMo PTY inheritance.
+//   The tool returns the log path as the durable handle.
+//
+// Wait-intent compilation (sleep/check/poll aimed at a pending operation)
+// is preserved: interception happens at the SEMANTIC level via
+// await_intent route-bash, and a NO-OP for anything that is not a genuine
+// continuation of a live pending operation. Arbitrary shell syntax
+// (quotes, pipes, semicolons) passes through untouched — the router
+// response is only trusted when it carries normalized=true AND
+// ContinuePendingOperation with a concrete operation bound.
+import { tool } from "@mimo-ai/plugin"
+import { spawn, execFile } from "child_process"
+
+const REPO = "/mnt/workdrive/issue161-harness"
+const WR_ROOT = "/mnt/workdrive/wr-bridge"
+
+// ── MODE A: process-group-owned capture execution ──────────────────────
+interface OwnedResult { stdout: string; stderr: string; status: number; timedOut: boolean }
+
+function runOwned(command: string, timeoutMs: number): Promise<OwnedResult> {
+    return new Promise((resolve) => {
+        // /bin/sh -c "command" in its OWN process group (detached:true makes
+        // the child a session/group leader we can signal as a WHOLE).
+        const child = spawn("/bin/sh", ["-c", command], {
+            cwd: REPO,
+            stdio: ["ignore", "pipe", "pipe"],   // NEVER inherit the TUI tty
+            env: process.env,
+            detached: true,                      // own process group
+        })
+        let stdout = "", stderr = ""
+        let timedOut = false
+        let settled = false
+        const outChunks: Buffer[] = []
+        const errChunks: Buffer[] = []
+        child.stdout!.on("data", (d) => outChunks.push(d))
+        child.stderr!.on("data", (d) => errChunks.push(d))
+
+        const killGroup = (sig: NodeJS.Signals) => {
+            try { process.kill(-child.pid!, sig) } catch { /* already gone */ }
+        }
+        const finalize = (status: number) => {
+            if (settled) return
+            settled = true
+            clearTimeout(timer)
+            // detach listeners, then resolve — no late listeners remain
+            child.stdout!.removeAllListeners()
+            child.stderr!.removeAllListeners()
+            resolve({ stdout: Buffer.concat(outChunks).toString(), stderr: Buffer.concat(errChunks).toString(), status, timedOut })
+        }
+        const timer = setTimeout(() => {
+            timedOut = true
+            killGroup("SIGTERM")
+            // bounded grace for clean shutdown, then SIGKILL the group
+            setTimeout(() => killGroup("SIGKILL"), 2000)
+            // the exit/close event settles; a hard stop-guard settles anyway
+            setTimeout(() => finalize(124), 4000)
+        }, timeoutMs)
+
+        child.on("error", (e) => finalize(-1))
+        child.on("close", (code) => finalize(code ?? -1))
+    })
+}
+
+// ── MODE B: explicit detachment for "cmd &" invocations ────────────────
+function runDetached(command: string): Promise<string> {
+    return new Promise((resolve) => {
+        const logPath = `/tmp/mimo-detached-${Date.now()}-${Math.floor(Math.random() * 1e6)}.log`
+        const out = `exec >>"${logPath}" 2>&1; exec 0</dev/null; ${command}`
+        const child = spawn("/bin/sh", ["-c", `setsid /bin/sh -c '${out.replace(/'/g, `'\\''`)}'`], {
+            cwd: REPO,
+            stdio: "ignore",     // child rebinds its own fds — no PTY inheritance
+            env: process.env,
+            detached: true,
+            // unref: MiMo's event loop does not wait for it
+        })
+        child.unref()
+        resolve(JSON.stringify({
+            detached: true,
+            log: logPath,
+            note: "process detached (setsid, own fds); output streams to the log file only; fd 0/1/2 do NOT reference the MiMo tty",
+        }))
+    })
+}
+
+// ── wait-intent routing (semantic level, preserved from v1) ────────────
+async function tryCompileWaitIntent(command: string): Promise<string | null> {
+    let routeOut: string
+    try {
+        routeOut = await new Promise<string>((resolve, reject) => {
+            execFile("python3", [`${REPO}/scripts/await_intent_cli.py`, "route-bash", "--command", command], {
+                cwd: REPO, timeout: 60000, encoding: "utf-8",
+                env: { ...process.env, RUNE_WR_ROOT: WR_ROOT },
+                maxBuffer: 8 * 1024 * 1024,
+            }, (err, stdout) => { if (err && !stdout) reject(err); else resolve(stdout || "") })
+        })
+    } catch {
+        return null // router unavailable: normal execution (never block on routing)
+    }
+    const jl = routeOut.trim().split("\n").find((l) => l.startsWith("{"))
+    if (!jl) return null
+    let norm: any
+    try { norm = JSON.parse(jl) } catch { return null }
+    // STRICT gate: compile ONLY a normalized, concrete continuation.
+    if (!(norm.normalized === true &&
+          norm.action === "ContinuePendingOperation" &&
+          norm.operation_id && norm.sha_for_wait)) return null
+    const waitCmd = `python3 ${REPO}/scripts/ci_event_wait.py --sha ${norm.sha_for_wait} --deadline ${norm.deadline_secs} --op-type ${norm.op_type}`
+    const r = await runOwned(waitCmd, (norm.deadline_secs + 60) * 1000)
+    const jline = r.stdout.trim().split("\n").find((l) => l.startsWith("{"))
+    if (jline) {
+        return JSON.stringify({
+            ...JSON.parse(jline),
+            intent_compilation: {
+                raw_command_kind: norm.raw_command_kind || "sleep_poll",
+                literal_sleep_executed: false,
+                operation_id: norm.operation_id,
+                canonical_op_type: norm.op_type,
+                authority: norm.authority,
+            },
+        })
+    }
+    return JSON.stringify({ ok: false, stage: "compiled-wait", error: (r.stderr || "no output").slice(0, 300) })
+}
+
+export default tool({
+    description: "Run a shell command. If the command expresses external wait intent (sleep/check/poll aimed at CI, deploy, or release operations with pending operations), it is automatically compiled to Rune's event-driven continuation — you do not need to manage wait mechanics. Plain local commands execute normally. Commands ending in ' &' run detached with output logged to a file.",
+    args: {
+        command: tool.schema.string().describe("Shell command to run"),
+    },
+    async execute(args) {
+        // 1. wait-intent compilation (semantic, strict-gated)
+        try {
+            const compiled = await tryCompileWaitIntent(args.command)
+            if (compiled !== null) return compiled
+        } catch { /* fall through to normal execution */ }
+
+        // 2. MODE B: explicit detachment for backgrounded invocations
+        const cmd = args.command
+        if (/\b(nohup|setsid)\b/.test(cmd) || /&\s*$/.test(cmd)) {
+            return await runDetached(cmd)
+        }
+
+        // 3. MODE A: tool-owned, group-killed, captured execution
+        const r = await runOwned(cmd, 120000)
+        const text = r.stdout + (r.stderr ? `\n[stderr]\n${r.stderr}` : "")
+        if (r.timedOut) {
+            return `exit timed-out (process group terminated): ${text.slice(0, 500)}`
+        }
+        if (r.status !== 0) {
+            return `exit ${r.status}: ${text.slice(0, 500) || "?"}`
+        }
+        return text || "(no output)"
+    },
+})
@@ -1,223 +1,272 @@
 {
-  "_comment": "Umbrella impact map \u2014 the change-propagation graph. Every src/rune/*.py and shared/*.py module MUST appear in exactly one umbrella's modules list (enforced by the umbrella_manifest_is_complete henxel; debt-tracked modules allowed via docs/module-coverage-debt.md). depends_on: when umbrella U changes, its dependents' canaries are required gates. mirrors: files that must travel together in the same PR (enforced by scripts/umbrella_impact.py unless the PR carries 'Propagation-waived: <reason>').",
-  "umbrellas": {
-    "memory": {
-      "description": "Storing, retrieving, and managing knowledge",
-      "modules": [
-        "memory_store",
-        "memory_consolidation",
-        "memory_env",
-        "shard_manager",
-        "query_memory",
-        "graph_memory",
-        "warm_neuron_store",
-        "abstraction",
-        "cache",
-        "recall_router",
-        "drift_cache",
-        "drift_compression",
-        "tier_manager",
-        "graph_shard_bridge",
-        "memory_selector",
-        "retrieval_benchmark"
-      ],
-      "canaries": [
-        "canary_query_memory",
-        "canary_memory_synthesis",
-        "canary_full_system"
-      ],
-      "depends_on": [
-        "governance"
-      ]
-    },
-    "governance": {
-      "description": "Quality control, drift detection, promotion, audit",
-      "modules": [
-        "drift_core",
-        "drift_observability",
-        "drift_ledger",
-        "audit_ledger",
-        "curator",
-        "claim_audit",
-        "contracts",
-        "build_pipeline_policy",
-        "failure_auto_filer"
-      ],
-      "canaries": [
-        "canary_full_system",
-        "canary_schema_migration"
-      ],
-      "depends_on": []
-    },
-    "identity": {
-      "description": "Who the agent is \u2014 persona, voice, anchors",
-      "modules": [
-        "persona_definition",
-        "persona_gate",
-        "identity_anchors",
-        "voice_hybrid",
-        "voice_sync",
-        "voicebox",
-        "voice_injector",
-        "behavior",
-        "prompt_sync"
-      ],
-      "canaries": [
-        "canary_full_system"
-      ],
-      "depends_on": [
-        "governance"
-      ]
-    },
-    "growth": {
-      "description": "Learning and improvement",
-      "modules": [
-        "self_improve",
-        "subconscious_pass",
-        "goal",
-        "elastic_cognition",
-        "growth_delta",
-        "autonomy_loop",
-        "cognitive_loop",
-        "dream_cycle",
-        "growth_simulator"
-      ],
-      "canaries": [
-        "canary_full_system"
-      ],
-      "depends_on": [
-        "memory",
-        "governance"
-      ]
-    },
-    "perception": {
-      "description": "Input processing and context assembly",
-      "modules": [
-        "context_broker",
-        "context_pressure",
-        "attention_capsule",
-        "auto_context",
-        "auto_tasks",
-        "witness",
-        "salience_gate",
-        "cognitive_budgeter",
-        "complexity_scorer"
-      ],
-      "canaries": [
-        "canary_context_broker",
-        "canary_full_system"
-      ],
-      "depends_on": [
-        "memory"
-      ]
-    },
-    "capability": {
-      "description": "Tool selection and execution authority",
-      "modules": [
-        "capability_broker",
-        "tool_executor",
-        "tool_protocol",
-        "local_tool_runtime",
-        "forge_client",
-        "message_queue",
-        "connection_pool",
-        "adaptive_router",
-        "route_risk_flags",
-        "evidence_probe",
-        "builder_bridge",
-        "builder_bridge_cli",
-        "recall_uplift"
-      ],
-      "canaries": [
-        "canary_p1_01",
-        "canary_p0_05",
-        "canary_native"
-      ],
-      "depends_on": [
-        "memory"
-      ]
-    },
-    "inference": {
-      "description": "Model routing and outbound requests",
-      "modules": [
-        "model_routing",
-        "model_profiles",
-        "model_registry",
-        "model_connect",
-        "provider_registry",
-        "provider_budget",
-        "standardized_providers",
-        "inference_trace",
-        "response_adapter",
-        "token_optimizer"
-      ],
-      "canaries": [
-        "canary_p1_06",
-        "canary_p1_02",
-        "canary_resource_authority"
-      ],
-      "depends_on": [
-        "capability"
-      ]
-    },
-    "orchestration": {
-      "description": "Task lifecycle, operator, TUI presentation",
-      "modules": [
-        "rune_operator",
-        "rune_tui",
-        "rune_graphic",
-        "rune_spinner",
-        "rune_profiler",
-        "chat_bubble",
-        "message_widgets",
-        "tui_validator",
-        "microkernel",
-        "event_bus",
-        "event_envelope",
-        "hook_engine",
-        "system_registry",
-        "cron_engine",
-        "task_gate",
-        "task_persist",
-        "work_manager",
-        "parallel_processor",
-        "tiered_executor",
-        "time_awareness",
-        "night_mode",
-        "pins",
-        "progressive_escalation",
-        "safety_override",
-        "sleep_time",
-        "verification_layer",
-        "outcome_ledger",
-        "warm_router",
-        "subsystem_health",
-        "build_mode",
-        "dependency_tracker",
-        "auto_detect",
-        "build_orchestrator",
-        "custodian",
-        "gitea_auth",
-        "gitea_pipeline",
-        "gitea_provision",
-        "umbrella_update"
-      ],
-      "canaries": [
-        "canary_p1_07",
-        "canary_orchestration_foundation",
-        "canary_conversation_state",
-        "canary_gitea_pipeline",
-        "canary_tui_drift",
-        "canary_cognitive_strata",
-        "canary_surface_parity"
-      ],
-      "depends_on": [
-        "perception",
-        "capability",
-        "inference",
-        "identity",
-        "growth"
-      ]
-    }
+ "_comment": "GENERATED by rune.spine \u2014 compatibility view for legacy umbrella consumers. DO NOT EDIT: structural truth lives in store/spine/current.json; edits here are flagged by spine drift.",
+ "_generated_by_spine": true,
+ "generated_at": "2026-09-16T18:23:38Z",
+ "umbrellas": {
+  "capability": {
+   "description": "Tool selection and execution authority",
+   "modules": [
+    "adaptive_router",
+    "builder_bridge",
+    "builder_bridge_cli",
+    "capability_broker",
+    "connection_pool",
+    "evidence_probe",
+    "forge",
+    "forge_client",
+    "intent_continuity",
+    "local_tool_runtime",
+    "message_queue",
+    "openai_compat_transport",
+    "provider_context",
+    "recall_uplift",
+    "route_risk_flags",
+    "tool_executor",
+    "tool_protocol"
+   ],
+   "depends_on": [
+    "memory"
+   ],
+   "mirrors": [],
+   "canaries": [
+    "canary_native",
+    "canary_p0_05",
+    "canary_p1_01"
+   ]
+  },
+  "governance": {
+   "description": "Quality control, drift detection, promotion, audit",
+   "modules": [
+    "audit_ledger",
+    "build_pipeline_policy",
+    "claim_audit",
+    "contracts",
+    "curator",
+    "drift_core",
+    "drift_ledger",
+    "drift_observability",
+    "failure_auto_filer",
+    "programme_gates"
+   ],
+   "depends_on": [],
+   "mirrors": [],
+   "canaries": [
+    "canary_full_system",
+    "canary_schema_migration"
+   ]
+  },
+  "growth": {
+   "description": "Learning and improvement",
+   "modules": [
+    "autonomy_loop",
+    "cognitive_loop",
+    "dream_cycle",
+    "elastic_cognition",
+    "goal",
+    "growth_delta",
+    "growth_simulator",
+    "self_improve",
+    "subconscious_pass"
+   ],
+   "depends_on": [
+    "governance",
+    "memory"
+   ],
+   "mirrors": [],
+   "canaries": [
+    "canary_full_system"
+   ]
+  },
+  "identity": {
+   "description": "Who the agent is \u2014 persona, voice, anchors",
+   "modules": [
+    "behavior",
+    "identity_anchors",
+    "persona_definition",
+    "persona_gate",
+    "prompt_sync",
+    "voice_hybrid",
+    "voice_injector",
+    "voice_sync",
+    "voicebox"
+   ],
+   "depends_on": [
+    "governance"
+   ],
+   "mirrors": [],
+   "canaries": [
+    "canary_full_system"
+   ]
+  },
+  "inference": {
+   "description": "Model routing and outbound requests",
+   "modules": [
+    "inference_trace",
+    "model_connect",
+    "model_profiles",
+    "model_registry",
+    "model_routing",
+    "provider_budget",
+    "provider_registry",
+    "response_adapter",
+    "standardized_providers",
+    "token_optimizer"
+   ],
+   "depends_on": [
+    "capability"
+   ],
+   "mirrors": [],
+   "canaries": [
+    "canary_p1_02",
+    "canary_p1_06",
+    "canary_resource_authority"
+   ]
+  },
+  "memory": {
+   "description": "Storing, retrieving, and managing knowledge",
+   "modules": [
+    "abstraction",
+    "cache",
+    "drift_cache",
+    "drift_compression",
+    "graph_memory",
+    "graph_shard_bridge",
+    "memory_consolidation",
+    "memory_env",
+    "memory_selector",
+    "memory_store",
+    "query_memory",
+    "recall_router",
+    "retrieval_benchmark",
+    "shard_manager",
+    "tier_manager",
+    "warm_neuron_store"
+   ],
+   "depends_on": [
+    "governance"
+   ],
+   "mirrors": [],
+   "canaries": [
+    "canary_full_system",
+    "canary_memory_synthesis",
+    "canary_query_memory"
+   ]
+  },
+  "orchestration": {
+   "description": "Task lifecycle, operator, TUI presentation",
+   "modules": [
+    "auto_detect",
+    "build_mode",
+    "build_orchestrator",
+    "chat_bubble",
+    "cron_engine",
+    "custodian",
+    "dependency_tracker",
+    "event_bus",
+    "event_envelope",
+    "gitea_auth",
+    "gitea_pipeline",
+    "gitea_provision",
+    "hook_engine",
+    "message_widgets",
+    "microkernel",
+    "night_mode",
+    "outcome_ledger",
+    "parallel_processor",
+    "pins",
+    "progressive_escalation",
+    "rune_graphic",
+    "rune_operator",
+    "rune_profiler",
+    "rune_spinner",
+    "rune_tui",
+    "safety_override",
+    "sleep_time",
+    "subsystem_health",
+    "system_registry",
+    "task_gate",
+    "task_persist",
+    "tiered_executor",
+    "time_awareness",
+    "tui_validator",
+    "umbrella_update",
+    "verification_layer",
+    "warm_router",
+    "work_manager"
+   ],
+   "depends_on": [
+    "capability",
+    "growth",
+    "identity",
+    "inference",
+    "perception"
+   ],
+   "mirrors": [],
+   "canaries": [
+    "canary_cognitive_strata",
+    "canary_conversation_state",
+    "canary_gitea_pipeline",
+    "canary_orchestration_foundation",
+    "canary_p1_07",
+    "canary_surface_parity",
+    "canary_tui_drift"
+   ]
+  },
+  "perception": {
+   "description": "Input processing and context assembly",
+   "modules": [
+    "attention_capsule",
+    "auto_context",
+    "auto_tasks",
+    "behaviour_ledger",
+    "build_basis",
+    "capability_fit",
+    "capability_ledger",
+    "cognitive_budgeter",
+    "construction_activation",
+    "construction_commands",
+    "construction_decision",
+    "construction_depth",
+    "construction_intent",
+    "construction_interface",
+    "construction_mode",
+    "construction_pattern",
+    "construction_patterns_builtin",
+    "construction_projection",
+    "context_broker",
+    "context_pressure",
+    "drift_resume",
+    "event_wait",
+    "failure_recovery",
+    "graph_projection",
+    "impact_preview",
+    "iteration",
+    "layout_contract",
+    "programme_continuation",
+    "project_registry",
+    "release_notes",
+    "salience_gate",
+    "scaffold",
+    "spine",
+    "stage_guards",
+    "structural_migration",
+    "structural_model",
+    "structural_store",
+    "testing_mode",
+    "toolchain",
+    "tui_backlog",
+    "wait_guard",
+    "wait_guard_corpus",
+    "witness"
+   ],
+   "depends_on": [
+    "memory"
+   ],
+   "mirrors": [],
+   "canaries": [
+    "canary_context_broker",
+    "canary_full_system"
+   ]
  }
+ }
 }
@@ -0,0 +1,225 @@
+{
+  "_comment": "Umbrella impact map \u2014 the change-propagation graph. Every src/rune/*.py and shared/*.py module MUST appear in exactly one umbrella's modules list (enforced by the umbrella_manifest_is_complete henxel; debt-tracked modules allowed via docs/module-coverage-debt.md). depends_on: when umbrella U changes, its dependents' canaries are required gates. mirrors: files that must travel together in the same PR (enforced by scripts/umbrella_impact.py unless the PR carries 'Propagation-waived: <reason>').",
+  "umbrellas": {
+    "memory": {
+      "description": "Storing, retrieving, and managing knowledge",
+      "modules": [
+        "memory_store",
+        "memory_consolidation",
+        "memory_env",
+        "shard_manager",
+        "query_memory",
+        "graph_memory",
+        "warm_neuron_store",
+        "abstraction",
+        "cache",
+        "recall_router",
+        "drift_cache",
+        "drift_compression",
+        "tier_manager",
+        "graph_shard_bridge",
+        "memory_selector",
+        "retrieval_benchmark"
+      ],
+      "canaries": [
+        "canary_query_memory",
+        "canary_memory_synthesis",
+        "canary_full_system"
+      ],
+      "depends_on": [
+        "governance"
+      ]
+    },
+    "governance": {
+      "description": "Quality control, drift detection, promotion, audit",
+      "modules": [
+        "drift_core",
+        "drift_observability",
+        "drift_ledger",
+        "audit_ledger",
+        "curator",
+        "claim_audit",
+        "contracts",
+        "build_pipeline_policy",
+        "failure_auto_filer"
+      ],
+      "canaries": [
+        "canary_full_system",
+        "canary_schema_migration"
+      ],
+      "depends_on": []
+    },
+    "identity": {
+      "description": "Who the agent is \u2014 persona, voice, anchors",
+      "modules": [
+        "persona_definition",
+        "persona_gate",
+        "identity_anchors",
+        "voice_hybrid",
+        "voice_sync",
+        "voicebox",
+        "voice_injector",
+        "behavior",
+        "prompt_sync"
+      ],
+      "canaries": [
+        "canary_full_system"
+      ],
+      "depends_on": [
+        "governance"
+      ]
+    },
+    "growth": {
+      "description": "Learning and improvement",
+      "modules": [
+        "self_improve",
+        "subconscious_pass",
+        "goal",
+        "elastic_cognition",
+        "growth_delta",
+        "autonomy_loop",
+        "cognitive_loop",
+        "dream_cycle",
+        "growth_simulator"
+      ],
+      "canaries": [
+        "canary_full_system"
+      ],
+      "depends_on": [
+        "memory",
+        "governance"
+      ]
+    },
+    "perception": {
+      "description": "Input processing and context assembly",
+      "modules": [
+        "context_broker",
+        "context_pressure",
+        "attention_capsule",
+        "auto_context",
+        "auto_tasks",
+        "witness",
+        "salience_gate",
+        "cognitive_budgeter",
+        "complexity_scorer"
+      ],
+      "canaries": [
+        "canary_context_broker",
+        "canary_full_system"
+      ],
+      "depends_on": [
+        "memory"
+      ]
+    },
+    "capability": {
+      "description": "Tool selection and execution authority",
+      "modules": [
+        "capability_broker",
+        "tool_executor",
+        "tool_protocol",
+        "local_tool_runtime",
+        "intent_continuity",
+        "openai_compat_transport",
+        "forge_client",
+        "message_queue",
+        "connection_pool",
+        "adaptive_router",
+        "route_risk_flags",
+        "evidence_probe",
+        "builder_bridge",
+        "builder_bridge_cli",
+        "recall_uplift"
+      ],
+      "canaries": [
+        "canary_p1_01",
+        "canary_p0_05",
+        "canary_native"
+      ],
+      "depends_on": [
+        "memory"
+      ]
+    },
+    "inference": {
+      "description": "Model routing and outbound requests",
+      "modules": [
+        "model_routing",
+        "model_profiles",
+        "model_registry",
+        "model_connect",
+        "provider_registry",
+        "provider_budget",
+        "standardized_providers",
+        "inference_trace",
+        "response_adapter",
+        "token_optimizer"
+      ],
+      "canaries": [
+        "canary_p1_06",
+        "canary_p1_02",
+        "canary_resource_authority"
+      ],
+      "depends_on": [
+        "capability"
+      ]
+    },
+    "orchestration": {
+      "description": "Task lifecycle, operator, TUI presentation",
+      "modules": [
+        "rune_operator",
+        "rune_tui",
+        "rune_graphic",
+        "rune_spinner",
+        "rune_profiler",
+        "chat_bubble",
+        "message_widgets",
+        "tui_validator",
+        "microkernel",
+        "event_bus",
+        "event_envelope",
+        "hook_engine",
+        "system_registry",
+        "cron_engine",
+        "task_gate",
+        "task_persist",
+        "work_manager",
+        "parallel_processor",
+        "tiered_executor",
+        "time_awareness",
+        "night_mode",
+        "pins",
+        "progressive_escalation",
+        "safety_override",
+        "sleep_time",
+        "verification_layer",
+        "outcome_ledger",
+        "warm_router",
+        "subsystem_health",
+        "build_mode",
+        "dependency_tracker",
+        "auto_detect",
+        "build_orchestrator",
+        "custodian",
+        "gitea_auth",
+        "gitea_pipeline",
+        "gitea_provision",
+        "umbrella_update"
+      ],
+      "canaries": [
+        "canary_p1_07",
+        "canary_orchestration_foundation",
+        "canary_conversation_state",
+        "canary_gitea_pipeline",
+        "canary_tui_drift",
+        "canary_cognitive_strata",
+        "canary_surface_parity"
+      ],
+      "depends_on": [
+        "perception",
+        "capability",
+        "inference",
+        "identity",
+        "growth"
+      ]
+    }
+  }
+}
@@ -104,15 +104,15 @@
        },
        {
          "path": "src/rune/attention_capsule.py",
-          "role": "AttentionCapsule — the typed ledger of admitted blocks the model receives (to_messages re-derives the runtime payload)",
+          "role": "AttentionCapsule \u2014 the typed ledger of admitted blocks the model receives (to_messages re-derives the runtime payload)",
          "disposition": "retain-authority",
-          "evidence": "P1-02 (#25): the capsule IS the product the model receives — the broker renders admitted blocks into it, the safety override mutates its user block in place (re-measured), runtime.run consumes capsule.to_messages(), and the trace v4 capsule_messages/capsule_content_hashes prove the logged capsule EXACTLY equals the sent payload. Legacy create_capsule/to_prompt (string-trim fake budgeting) deprecated with no live callers."
+          "evidence": "P1-02 (#25): the capsule IS the product the model receives \u2014 the broker renders admitted blocks into it, the safety override mutates its user block in place (re-measured), runtime.run consumes capsule.to_messages(), and the trace v4 capsule_messages/capsule_content_hashes prove the logged capsule EXACTLY equals the sent payload. Legacy create_capsule/to_prompt (string-trim fake budgeting) deprecated with no live callers."
        },
        {
          "path": "src/rune/inference_trace.py",
          "role": "Durable per-run inference trace (tokens served/estimated, output, latency, per-block visibility)",
          "disposition": "migrate-adapter",
-          "evidence": "P0-07 (#23): persists what ContextBroker (#19) assembled and what the transport served — builds FROM the broker trace, never re-measures; store state/inference_traces.jsonl."
+          "evidence": "P0-07 (#23): persists what ContextBroker (#19) assembled and what the transport served \u2014 builds FROM the broker trace, never re-measures; store state/inference_traces.jsonl."
        }
      ]
    },
@@ -160,13 +160,13 @@
          "path": "src/rune/capability_broker.py",
          "role": "Authoritative capability selector (select_for_job + verified floor + find_capability escape hatch)",
          "disposition": "retain-authority",
-          "evidence": "P1-01 (#24): select_for_job is the single selection authority — its ordered decision feeds LocalToolRuntime(tools=) at both construction sites and the _contributor_tools prompt block; degrade-not-crash falls back to full TOOLS with a tier-1 record_failure."
+          "evidence": "P1-01 (#24): select_for_job is the single selection authority \u2014 its ordered decision feeds LocalToolRuntime(tools=) at both construction sites and the _contributor_tools prompt block; degrade-not-crash falls back to full TOOLS with a tier-1 record_failure."
        },
        {
          "path": "memory/index/capability_cards.json",
          "role": "Capability-card index (rebuilt on empty/stale/corrupt; native cards synthesized from TOOLS metadata)",
          "disposition": "migrate-adapter",
-          "evidence": "P1-01 (#24): load_cards validates and rebuilds — the packaged empty array no longer passes as an index; cards carry risk/preconditions/side_effects copied from tool_executor.TOOLS."
+          "evidence": "P1-01 (#24): load_cards validates and rebuilds \u2014 the packaged empty array no longer passes as an index; cards carry risk/preconditions/side_effects copied from tool_executor.TOOLS."
        },
        {
          "path": "src/rune/local_tool_runtime.py",
@@ -486,12 +486,6 @@
          "disposition": "retain-authority",
          "evidence": "Defines the owner-approved canary-first Gitea sequence."
        },
-        {
-          "path": "src/rune/plugins/build_mode/__init__.py",
-          "role": "Checkpoint/ledger state-machine primitive used by current governed builds",
-          "disposition": "migrate-adapter",
-          "evidence": "Persists checkpoints but encodes an older phase order."
-        },
        {
          "path": "src/rune/build_mode.py",
          "role": "Separate BuildManager and worktree engine",
@@ -504,6 +498,9 @@
          "disposition": "archive-candidate",
          "evidence": "Reports fixed success strings for phases without executing their claimed effects."
        }
+      ],
+      "notes": [
+        "T17 17U: plugin BuildStateMachine deleted (zero importers; second parallel orchestrator)"
      ]
    },
    {
@@ -540,4 +537,4 @@
      ]
    }
  ]
-}
+}
@@ -0,0 +1,141 @@
+# Gitea Issue Reconciliation — canonical record
+
+_Generated T15 Part J/K/L (2026-09-11). Machine-readable companion:
+`docs/gitea_issue_reconciliation.json` (single authority; this doc is the
+human-readable projection). Consistency checker: `scripts/gitea_reconciliation_check.py`
+compares live Gitea state against the JSON and fails on drift._
+
+Reconciliation states (closed vocabulary):
+`VALID_OPEN | FIXED_VERIFIED | SUPERSEDED | DUPLICATE | HISTORICAL_EVIDENCE |
+ACCEPTED_LIMITATION | NEEDS_REVALIDATION | DEFERRED_FUTURE`
+
+## Bug-label audit (Part K)
+
+`bug` = demonstrated currently incorrect behavior. Nothing else.
+
+### Issues labelled `bug` at audit time
+
+| # | live defect | evidence | final label | reconciliation state |
+|---|---|---|---|---|
+| 324 | **YES** | MessageAbortedError reproduces every aborted worker session (15+ checkpoints, including this session's predecessor) | bug (kept) | VALID_OPEN |
+| 204 | NO | Forge-orchestrated pipeline = feature request (P1 feature); no demonstrated incorrect behavior | bug removed | VALID_OPEN (feature) |
+| 157 | NO | capture_failure.py + tests/fixtures/production/ + manifest.json shipped and exercised (plan 2026-09-07-157) | bug removed | FIXED_VERIFIED (closed) |
+| 145 | NO | P0-07 follow-up work item (tracing coverage), not a defect report | bug removed | VALID_OPEN (work item) |
+| 139 | NO | Hand-maintained allowlist replaced by declarative classification (tui_validator.py cites #139); `tui_validator.py check` → passed, exit 0 | bug removed | FIXED_VERIFIED (closed) |
+| 126 | NO | Silent-handler sweep = refactor program (S110/BLE001 ratchet governs), not a defect | bug removed | VALID_OPEN (work item) |
+| 37 | NO | Personality corpus compilation = P2 feature | bug removed | VALID_OPEN (feature) |
+| 16 | NO | v1 course-correction EPIC | bug removed | VALID_OPEN (epic) |
+
+### Issues whose `bug` label was removed in the prior cleanup (Part D, earlier session)
+
+16 targeted: #301 #300 #292 #284 #314 #181 #166 #165 #163 #159 #154 #141 #40 #32 #31 #27.
+Verification (live API, this session): all confirmed bug-free EXCEPT #314 — its
+removal had failed (the API-False defect) and the label persisted on a closed
+issue. Removed this session by label-ID DELETE (204). #314 is closed
+FIXED_VERIFIED (mode:association_merged regression shipped in 2bf43ea).
+
+### Totals
+
+- bug-labelled before audit: 8 open (+ #314 closed-but-labelled)
+- bug-labelled after audit: **1** (#324, live defect, evidence above)
+- issues closed by this audit: #139, #157 (both FIXED_VERIFIED with evidence comments)
+- NEEDS_REVALIDATION count: 0 among bug-labelled issues (each resolved on evidence)
+
+## Open-issue reconciliation (Part J)
+
+Subsystems derived from labels. States per closed vocabulary. `next action` is
+the single next concrete step; `_` = none until its tranche opens.
+
+| # | subsystem | state | current evidence | next action |
+|---|---|---|---|---|
+| 324 | engine/checkpoint | VALID_OPEN | live repro (15+ sessions) | fix writer abort handling |
+| 316 | ADAPT-1/research | VALID_OPEN | B3-v2 research task | await ADAPT tranche |
+| 315 | ADAPT-1/wiring | VALID_OPEN | B2-v2 standard task | await ADAPT tranche |
+| 311 | ADAPT-1/wiring | VALID_OPEN | superseded-by #315 (kept open as spec anchor until 315 delivers) | close when #315 closes |
+| 301 | FINAL-12/benchmarks | NEEDS_REVALIDATION | dependency list silently grew (operator-path telemetry, bridge overhead) without issue update | re-scope before PL1 metrics work |
+| 300 | ADAPT-11 | VALID_OPEN | acceptance = real escalation demo; blocked on ADAPT-LIVE telemetry | demo after PL1 held-out |
+| 299 | ADAPT-10 | VALID_OPEN | harness legibility audit | _ |
+| 297 | ADAPT-8 | VALID_OPEN | deterministic-first QUALIFY | _ |
+| 296 | ADAPT-7 | VALID_OPEN | DESIGN consolidation experiment | _ |
+| 295 | ADAPT-6 | VALID_OPEN | evidence capsule / context minimization | _ |
+| 294 | ADAPT-5 | VALID_OPEN | budgets + stop conditions | _ |
+| 293 | ADAPT-4 | VALID_OPEN | DESIGN-LITE shipped (STANDARD_CHANGE lane live) | acceptance demo, then close |
+| 292 | ADAPT-3 | VALID_OPEN | LOCAL_FIX lane shipped | acceptance demo, then close |
+| 284 | ADAPT epic | VALID_OPEN | epic tracks ADAPT-LIVE + PL1; body updated with detour record | update after PL1 |
+| 259 | Scaffolding | VALID_OPEN | redefined: safe repeatable construction protocol (T15 §C) | Scaffolding v1 after PL1 |
+| 258 | State-aware processes | VALID_OPEN | redefined: low-compute architectural self-awareness; SA1-8 superseded by distributed concerns | deferred research |
+| 204 | forge/pipeline | VALID_OPEN | P1 feature: forge-orchestrated coordination | _ |
+| 183–176 | memory-foundation futures | DEFERRED_FUTURE | P3 research futures (typed temporal semantics, prior-art audit, curiosity, lineage) | memory-foundation tranche |
+| 175/181 | memory epics | DEFERRED_FUTURE | evidence-grounded continuity epics | same |
+| 166 | research watchlist | DEFERRED_FUTURE | external research tracking | periodic review |
+| 165/164 | research spikes | DEFERRED_FUTURE | SimpleMem/A-MEM, HippoRAG 2 | memory-foundation tranche |
+| 163 | graph backend spike | NEEDS_REVALIDATION | P1 spike; LadybugDB evaluation partially done in prior research artifacts | revalidate against current graph.db-missing state |
+| 162 | Qdrant spike | NEEDS_REVALIDATION | P1 spike; prior research artifacts exist (issue-162/) | revalidate after #316 outcome |
+| 159 | conceptual topology | DEFERRED_FUTURE | P3 future | _ |
+| 158 | wiring-status report | VALID_OPEN | P2: generated architecture-truth report | _ |
+| 156 | mutation guards | VALID_OPEN | P1: cognitive-state mutation guards | _ |
+| 155 | e2e activation trace | VALID_OPEN | P1: credit-assignment trace | overlaps PL1 outcome work — potential duplicate of #300's telemetry; review at PL1 close |
+| 154/153 | learned-state / experience encoding | VALID_OPEN | P2 | PL1-adjacent |
+| 152/151 | break-it / adversarial | VALID_OPEN | P0: adversarial procedures after nominal green | post-PL1 tranche |
+| 145 | P0-07 follow-up | VALID_OPEN | tracing coverage gaps | _ |
+| 141 | ContextBroker compaction | VALID_OPEN | P1: rolling-compaction persistence | _ |
+| 139 | tui_validator drift | FIXED_VERIFIED | declarative classification shipped; check passes | closed |
+| 131/130/129 | reviewer program | VALID_OPEN | P0: local reviewer + calibration + rubric | post-PL1 tranche |
+| 126 | silent-handler sweep | VALID_OPEN | P0 refactor program; ratchet enforces ceiling | ongoing |
+| 125 | decompose monoliths | VALID_OPEN | P2 | _ |
+| 49–35 | v1-course-correction P2/P3 | DEFERRED_FUTURE | epic #16 manifest items (P2-01..09, P3-01..06) | per epic #16 sequencing |
+| 33–27 | v1-course-correction P1 | VALID_OPEN | epic #16 manifest P1 items | per epic #16 sequencing |
+| 16 | v1 epic | VALID_OPEN | consolidation epic; 33 manifest issues | per epic sequencing |
+| 13 | TUI message graph | VALID_OPEN | P1 | _ |
+| 3 | financial tracking | NEEDS_REVALIDATION | non-Rune issue (different project era) | transfer to correct repo or close |
+| 2 | cognitive loop wiring | NEEDS_REVALIDATION | predates current TUI architecture | revalidate against current TUI |
+| 1 | personality injection | NEEDS_REVALIDATION | overlaps #37's newer scope | merge into #37 or close |
+
+## Reconciliation decisions this session (Part K deltas)
+
+- CLOSED: #139, #157 (FIXED_VERIFIED, evidence comments posted)
+- BUG LABEL REMOVED: #204, #145, #126, #37, #16 (audit: not defects)
+- BUG LABEL REMOVED (stuck from prior cleanup): #314 (closed issue)
+- OPEN BUGS: 8 → 1 (#324 only — genuine live defect)
+
+## Part L — consistency check
+
+`scripts/gitea_reconciliation_check.py` (deterministic, no governance
+subsystem): fetches live issue state, compares to
+`docs/gitea_issue_reconciliation.json`, flags:
+
+- closed issue recorded VALID_OPEN / DEFERRED_FUTURE / NEEDS_REVALIDATION (open states)
+- open issue recorded FIXED_VERIFIED / SUPERSEDED / DUPLICATE / HISTORICAL_EVIDENCE without being closed
+- `bug` label present on any issue not recorded as a live defect (or missing on #324)
+- issue missing from the artifact
+- record lacking a required field
+
+Exit 0 = consistent; exit 1 = drift listed. CI-eligible (read-only).
+
+
+## PL1 held-out era reconciliation (2026-09-12)
+
+Fresh-session reconciliation of the ADAPT family against deployed 0.66.22
+@ 1497c85 + live PL1 held-out evidence (tasks task_2661189c9073,
+task_6e4885031355; ledger PE-9207213099-0270). Resolves the recorded tension:
+"ADAPT-LIVE complete → return target PL1" (the bounded tranche authorized
+2026-09-10 is complete) while numbered ADAPT issues remain open (the larger
+backlog was never in the ADAPT-LIVE tranche scope).
+
+| # | state | evidence / basis |
+|---|---|---|
+| 284 | VALID_OPEN | epic tracker for the remaining sub-items below |
+| 292 | FIXED_VERIFIED (closed) | LOCAL_FIX fast path live (rune_operator.py:522,541); live-proven twice by the held-out tasks |
+| 293 | FIXED_VERIFIED (closed) | DESIGN-LITE implemented (rune_operator.py:660, cites #293), wired to STANDARD_CHANGE; suite+canary green |
+| 294 | DEFERRED_FUTURE | canonical remaining-obligations list ("ADAPT-5 budgets... for later epics") |
+| 295 | DEFERRED_FUTURE | canonical remaining-obligations list |
+| 296 | DEFERRED_FUTURE | canonical remaining-obligations list |
+| 297 | FIXED_VERIFIED (closed) | task_gate implemented (src/rune/task_gate.py) + broker wiring + canary registration |
+| 299 | DEFERRED_FUTURE | canonical remaining-obligations list |
+| 300 | VALID_OPEN | escalation mechanism exists (_probe_and_escalate); fresh deployed acceptance evidence pending |
+| 301 | DEFERRED_FUTURE | canonical remaining-obligations list (FINAL-12) |
+| 315 | VALID_OPEN | not started; ADAPT-1 sub-item (no canonical deferral authority) |
+| 316 | VALID_OPEN | not started; ADAPT-1 sub-item (no canonical deferral authority) |
+
+Consistency checker: `scripts/gitea_reconciliation_check.py` → exit 0
+(72 records, 100 live issues in window) after these updates.
@@ -0,0 +1,53 @@
+# Rune Semantic Release Policy (T20 §20B)
+
+Pre-1.0 convention. Scaffolding "V1" does NOT mean Rune 1.0.
+
+## Bump kinds
+
+**PATCH** (0.66.80 → 0.66.81):
+- bug fix, correction, implementation/internal maintenance
+- no meaningful new user/system capability
+
+**MINOR** (0.66.80 → 0.67.0):
+- meaningful new Rune capability
+- new subsystem functionality
+- materially expanded system behaviour
+- significant new supported workflow
+
+**MAJOR**: reserved for Rune 1.0+ compatibility/public-contract semantics.
+
+## How the machinery expresses this (no manual version edits)
+
+The existing machinery already maps commit shape → bump kind:
+- `semver_bump.next_version` (PR titles): feat→minor, else patch
+- `version_convergence.highest_warranted_bump` (push ranges):
+  feat/breaking→minor, fix/perf→patch, else None
+
+**Tranche-level rule (20B)**: a feature tranche lands its feature
+commits (`feat(...)`), and the convergence machinery bumps MINOR at
+the tranche's release boundary. Patch-only follow-ups (fix/perf)
+warrant only PATCH — the machinery cannot accidentally re-minor
+because `highest_warranted_bump` takes the max over the *post-
+boundary* range, and a fix-only range yields "patch".
+
+## Why T20 → 0.67.0
+
+T20 is a feature-level tranche (Rune-wide retrofit: structural
+taxonomy, capability/behaviour ledgers, rune graph, /testing backlog,
+Spine modernization, semantic release). The release is classified
+MINOR under this policy, earned through the governed convergence
+machinery — never a manual version edit.
+
+## T20 0.67.0 release record
+T20 classified MINOR under this policy: Rune-wide retrofit (structural
+taxonomy, capability/behaviour ledgers, rune graph, /testing backlog,
+Spine modernization, semantic release) = new subsystem functionality +
+materially expanded system behaviour. Earned through the governed
+convergence machinery (feat-bearing range → minor), never a manual edit.
+
+## Root cause record (T20, 2026-09-16)
+semver_bump re-derived the bump kind from the SYNTHESIZED title text —
+"minor: converged release" matched nothing → patch. Feature tranches
+T18/T19/T20 downgraded to 0.66.77-83. Fixed: MINOR_PREFIX_RE carries the
+convergence kind through; first push after the fix with a feat-bearing
+range lands 0.67.0 through the normal machinery.
@@ -52,9 +52,11 @@ This detour is complete when:
 - [x] ADAPT intent recovered and reconciled (this artifact, Part B)
 - [ ] **THIS TRANCHE**: the ADAPT status matrix disposition is agreed and the next tranche selected (Part D)

-## RETURN TARGET
+## RETURN TARGET / CURRENT: PL1 Process Learning — FIRST TRANCHE LIVE (2026-09-11)

-**PL1 Process Learning** (T9). ADAPT-LIVE closed the route-execution gap; PL1's input (route accuracy, probe outcomes, escalation evidence, verification results, task characteristics) is now produced live by the operator spine. The matrix in the ADAPT RECONCILIATION section stands, updated: LOCAL_FIX/STANDARD_CHANGE paths = BUILT_AND_LIVE; probe = BUILT_AND_LIVE (verification-driven); escalation/downgrade = BUILT_AND_LIVE. Remaining ADAPT obligations for later epics: ADAPT-5 budgets, ADAPT-6 evidence capsule, ADAPT-7 DESIGN consolidation, ADAPT-10 legibility, FINAL-12 cross-track benchmark.
+PL1-01 structured Experience records: `scripts/pl1_records.py` (task_id, route resolved from route.jsonl, proof level, failure class, candidate status). PL1-02 failure classification in the PL0 vocabulary (research_dispatch, NOT_COMPLETED_NO_MUTATION_EVIDENCE, verification_failure, model_refusal, UNKNOWN). PL1-03 first learning loop: recurring prose-instead-of-mutation family → structural correction (route-driven instruction in the user context block) + replay contract (NOT_COMPLETED gate) — record PE-1789102658-0269 (SEED 2026-09-11 → VALIDATED 2026-09-12: two fresh deployed held-out mutation tasks completed-with-evidence — task_2661189c9073 + task_6e4885031355, LOCAL_FIX, Verboo-pinned, mutation + exit-code verification, ledger PE-9207213099-0270; substrate corrections 6965fc7 flat-ChatMessage contract + a2463ad resume transport selection applied between; caveat: corrections developer-applied structurally, not autonomous self-modification; PROMOTED requires repetition/soak). PL1-04 promotion discipline: SEED→CANDIDATE→VALIDATED→PROMOTED via process_experience; structural fixes preferred. PL1-05 metrics: first_pass_accepted, mutation_failures, research_dispatches, false_completion_rate (0.0 — structurally prevented), candidates vs evidence-only split.
+
+Remaining ADAPT obligations for later epics: ADAPT-5 budgets, ADAPT-6 evidence capsule, ADAPT-7 DESIGN consolidation, ADAPT-10 legibility, FINAL-12 cross-track benchmark, D7 release-job-on-Pi.

 ## OPEN OBLIGATIONS (deferred defects — recorded, NOT chased)

@@ -144,3 +146,44 @@ Reasoning from recovered obligations, not momentum:
 5. Update epic #284 with the detour record + this matrix; close the gap between tracker and reality.

 Then: **PL1** (T9), with the intent-continuity spine as PL1's first or second deliverable (it is an experience-system product, not a separate architecture project).
+
+## GITEA RECONCILIATION (2026-09-11)
+
+Part A-J executed. 15 issues closed (HISTORICAL_EVIDENCE/SUPERSEDED/ACCEPTED_LIMITATION) with evidence. Bug label cleaned: 16 misleading labels removed (open bugs 24→8). #259/#258 updated to current architecture. Remote tracking fixed (master tracks gitea/master, not stale origin). PL1 metrics corrected (NOT_YET_MEASURABLE where evidence doesn't support numbers). D7: bump works, release job fails on Pi — ROOT_CAUSE_UNKNOWN_DEFERRED. Automation noise: 5 duplicate bump commits per logical change (queue-drain artifact, not live recursion — version converges, commit history doesn't).
+
+## DEFECT D7 (2026-09-11): RESOLVED — release-job failures root-caused
+
+History: the release job failed on every push since 03:16 (c2528c1).
+Superseded diagnosis: "Pi-side backup step logs Failed to connect to
+127.0.0.1:3000" was a red herring — the Pi backup step was never the
+failing job.
+
+Root causes (live-proven T15 Part F, 2026-09-11 evening):
+1. c2528c1 left a corrupted shell fragment after the converged-check fi
+   in auto-version.yml — the release job died at bash parse time (exit 2)
+   before any bump. Fixed in c35b604.
+2. release_convergence() never assigned the release boundary — all 959+
+   commits classified as unreleased, so the warranted bump never
+   converged and queued runs each wrote their own bump commit. Fixed in
+   e395ee1 (boundary wiring) + 9e0b855 (boundary grep must match subjects
+   only, not bodies) + c35018f (numeric semver comparison: '0.66.9' <
+   '0.66.10' is False lexicographically, which silently skipped warranted
+   bumps as version_ahead_of_target).
+3. The per-file contents PUT committed one commit per version surface —
+   five same-version bump commits per real bump (the "duplicate bump
+   cascade" since 0.66.2, misdiagnosed as queue drain). Fixed in 470d82f
+   + d68fd1e (single ChangeFiles commit).
+
+Live proof: experiment A (trigger d68fd1e, warranted patch) → exactly one
+bump commit a4afc77 (0.66.11), all five surfaces in one commit, queue
+drained, zero delayed duplicates, zero workflow failures.
+
+## Local CI tooling (developer-side, NOT Rune runtime): W/R-v0.3 reactive wake (2026-09-12)
+
+The desktop gitea-runner (v3.3.0) now runs `host.post_task_script` = `/usr/local/bin/gitea-wr-wake.sh`
+after every job: a sub-second fire-and-forget POST to the local W/R daemon (`127.0.0.1:8791/wake`),
+which sweeps waiting external waits through AUTHORITATIVE Gitea verification (bounded retry for the
+reporter-ack window; 60s reconciler demoted to fallback). Runner event ≠ success proof. Config:
+`/etc/gitea-desktop-runner/config.yaml` (backup: config.yaml.pre-wr03). Measured: terminal→verified
+wait transition ~2s (was 0-30s polling). INTERACTIVE_AUTO_RESUME remains NOT YET SAFE (no supported
+MiMo busy/idle signal) — headless-only dispatch, unchanged.
@@ -0,0 +1,275 @@
+# Scaffolding V1 — Tranche 1
+
+_2026-09-12. Scaffolding is the controlled need-to-capability system: NEED →
+DISCOVERY → GAP → REUSE/ACQUIRE/BUILD → IMPACT+PLAN → SAFE IMPLEMENTATION →
+VERIFICATION → PROVISIONAL USE → TRUST/REVISE/RETIRE → REUSABLE CAPABILITY.
+Tranche 1 is deliberately narrower: make Rune stop rediscovering how Rune is
+built every time Rune changes._
+
+## Tranche 1 contract
+
+`scripts/scaffold_projection.py` — a PROJECTION over existing authorities,
+never a new one:
+
+| question | authority projected |
+|---|---|
+| what subsystem owns X? | `config/umbrellas.json` (umbrella manifest + henxel-enforced completeness) |
+| what does X depend on? | AST import graph over `src/rune/*.py` (incl. relative/lazy imports) |
+| who consumes X? | reverse import edges |
+| what changes if X changes? | mirrors + `depends_on` umbrellas + consumers |
+| what proof obligations arise? | unit-test file, canary registry, umbrella gates, suite/lint/henxels |
+| which are satisfied? | filesystem + registry evidence (static); suite/ratchet/henxels are runtime-check commands |
+
+Commands: `component <module>` · `impact <module>` · `obligations <module>` ·
+`all [outfile]`. If the projection disagrees with an authority, fix the
+authority — never the projection.
+
+## Mechanism audit (2026-09-12, deployed 0.66.22)
+
+| mechanism | verdict | basis |
+|---|---|---|
+| Umbrella manifest + impact map + propagation gate | KEEP | ownership/impact authority; henxel-enforced completeness; consumed by projection |
+| Henxels structural contract | KEEP | cross-cutting obligations (silent-handler ratchet, registries, coverage debt) |
+| Canary registry (`canary_native.MODULES` + per-track canaries) | KEEP | per-subsystem proof obligations; consumed by projection |
+| CI/deploy contract (required checks, TUI startup canary, atomic promote) | KEEP | deployment proof chain live-proven repeatedly |
+| Version surfaces + auto-version + semver | KEEP | bump fast lane + push convergence working (0.66.22 earned this session) |
+| Lint/pyright/coverage ratchets | KEEP | shrinking-baseline gates, all green |
+| Authority map + dead-subnet ledger (D6) | EVOLVE | static doc today; Scaffolding projections make it derivable; D6 retire-disposition still owed |
+| Issue reconciliation artifact + checker | KEEP | tracker truth with drift detection |
+| PL1 experience ledger | KEEP | the trust/revise/retire input for future capability crystallisation |
+| Intent continuity ledger | KEEP | PART 6 state machine, 7 invariants |
+| `MIMO_AUTOMATIC_RESUME` / W/R interactive injection | KEEP (bounded) | headless single-flight proven; interactive auto-resume NOT YET SAFE until a supported busy signal exists |
+| Duplicate authorities risk | REPLACE (avoided) | Scaffolding v1 adds NO new registry — projections only |
+
+## Real-change proof (this tranche)
+
+Change: context discovery hit Ollama's `/api/show` with cloud-pinned model
+names — `400 invalid model name` noise on every cloud task + 8192 misbudget
+for a 131072-window model (observed live in every PL1 Task A/B run).
+
+1. BEFORE mutation: `scaffold_projection.py impact local_tool_runtime`
+   derived — subsystem `capability`; consumers `rune_operator`; gates:
+   full suite, lint ratchet, henxels, inference+orchestration canaries;
+   unit-test obligation `tests/test_local_tool_runtime.py` (exists).
+2. Change made: `is_cloud_pinned` — THE provider-prefix rule, single
+   authority in `openai_compat_transport.py`; consumed by context discovery
+   (new) and transport selection (deduped); red-first regression tests.
+3. Obligations tracked: suite 1314 green · lint ratchet PASS · henxels exit 0
+   · the projection itself caught two gaps mid-flight (relative imports
+   hidden consumers; a broad except biting the lint ratchet) — both fixed at
+   the source the projection pointed to.
+4. Deploy: normal release chain → deployed identity verified.
+
+## Maturity (honest)
+
+- Projection engine: **UNIT-PROVEN** + proven on one real change.
+- Not yet: CI integration of `scaffold_projection.py all` (drift detection),
+  capability discovery / gap recognition / reuse-before-build / external
+  acquisition / provisional-trust lifecycle — FOUNDATIONAL ONLY (architecture
+  does not block them; the projection source-list is the extension surface).
+- Deferred with reasons recorded: none blocked by Tranche 1.
+
+---
+
+# Scaffolding V1 — Tranche 2: Construction Substrate (2026-09-13)
+
+_Tranche 2 (contract SCAFFOLD-TRANCHE-001) built the construction substrate
+itself: a shared structural-store primitive, the Project Scaffold layer over
+it, and durable Construction Contracts — Rune's first build tracked by the
+system it builds._
+
+## Architecture
+
+    src/rune/structural_store.py   shared generic primitive — NO Rune constants
+            ↓                               ↓
+    src/rune/spine.py            src/rune/scaffold.py
+    (Rune Spine — canonical      (Project Scaffold — canonical
+     structural self-model)       construction/structural truth
+                                  about EXTERNAL projects)
+
+Authority map (no duplicates):
+
+| fact | authority | others |
+|---|---|---|
+| Rune's structure | `spine.py` + `store/spine/current.json` | scaffold only references |
+| external project structure | `scaffold.py` store | Spine holds bridge facts only |
+| content history / rollback | Git | both reference commits |
+| coordination (issues/PRs/CI) | Gitea | scaffold stores observations |
+| why work exists | Task/Intent | contracts reference task ids |
+
+Construction Contract = ONE durable object (`contract:<id>`, kind
+`construction_contract`) with plain `phase`/`status` fields governed by a
+small legal-transition map (draft → approved → building → verifying →
+migrating → superseding → closed; failed/abandoned terminal-or-retry). No
+workflow engine. Resumability = store durability, proven across real
+process boundaries.
+
+## Mechanics
+
+- **Repo binding** (`bind_repo`): forge/owner/name/branch/root/placement as
+  scaffold DATA. No secrets — schema rejects secret-named fields AND every
+  `save()` recursively redacts credential-named keys (defense in depth in
+  the shared core, so Spine inherits it too).
+- **Scope check** (`check_scope`): changed files vs approved scope paths →
+  green / red / **unknown** (an unscoped contract is never silently green).
+  Out-of-scope change = RED until explicitly amended; amendments recorded,
+  original RED preserved in history.
+- **Predicted vs observed**: frontier snapshot at approval; observed impact
+  captured at closure from INDEPENDENT evidence (git, tests) — never from
+  re-walking the prediction's own graph. Missed entities become CANDIDATE
+  corrections awaiting evidence.
+- **Drift**: authority-owned facts that fail re-observation go STALE —
+  never silently fresh, never force-reverted. Reality wins.
+- **Bridge** (`spine_bridge_entity`): identity + repo surface only. Project
+  internals NEVER leak into Rune Spine (test-enforced). Contract evidence
+  may cite `spine://` URIs — citations are inert on the referenced
+  authority.
+
+## Real proof (this tranche on itself)
+
+SCAFFOLD-TRANCHE-001 migrated from its pre-scaffold JSON into the new
+store, advanced through phases with evidence checkpoints, closed with
+predicted-vs-observed from git/test evidence, scope check caught the
+store's own runtime artifacts (RED → explicit amendment → green, history
+kept), and construction provenance attached into Rune Spine without
+duplicating authority. A fresh interpreter answers what/why/state/
+evidence/changed/next from the store alone — no transcripts.
+
+## Maturity (honest)
+
+| component | state |
+|---|---|
+| structural_store core | EXERCISED (unit-proven + real use) |
+| Rune Spine after refactor | LIVE_PROVEN behavior-neutral (68 tests + identical rebuild) |
+| Project Scaffold | EXERCISED |
+| Construction Contract | EXERCISED (resumability proven cross-process) |
+| repo binding (identity model) | UNIT-PROVEN (local git fixture) |
+| external-repo live proof | BLOCKED ON USER REPO SELECTION |
+| Scaffolding overall | substrate only — capability-fit discovery, gap decisions, toolchain selection, patterns, maintenance TUI all NOT BUILT (data seams exist) |
+
+---
+
+# Scaffolding Tranche 4 — Live Need → Decision → Construction Contract (2026-09-13)
+
+_The adoption tranche: a real Need now flows automatically through the
+Tranche 2/3 mechanisms and produces a governed Construction Contract. The
+model no longer has to remember the procedure — it is durable Scaffolding._
+
+## Live flow
+
+    NEED (reference to Task/Intent — pointer, never a copy)
+      ↓ normalization (gaps recorded, never invented)
+      ↓ structural discovery + fit + resolution (Spine + Project Scaffold)
+      ↓ toolchain decision (existing project toolchain wins)
+      ↓ layout obligations (USER-LOCKED authority inherited)
+      ↓ user-approval gate (material changes require approval; approved
+        bounded work never re-asks)
+      ↓ DecisionTrace (durable, model-independent)
+      ↓ generated Construction Contract → existing build/verify pipeline
+
+## Authority map (unchanged — the path CONNECTS, never duplicates)
+
+| authority | owns |
+|---|---|
+| Task/Intent | why work exists |
+| Rune Spine | Rune structural truth |
+| Project Scaffold | external-project structural/construction truth |
+| Layout Contract | desired interface constraints (USER-LOCKED = user authority) |
+| Construction Contract | governed build obligations/history |
+| Git / Gitea | history / execution |
+
+## Boundary honesty
+
+- **ADAPT/research**: the live path records a research ROUTE
+  (TARGETED_EVIDENCE_PROBE / FULL_RESEARCH_PIPELINE) when evidence is
+  insufficient — a seam, NOT an executed pipeline call. No fake
+  integration claimed.
+- **BUILD_NEW** is never a fallback: insufficient evidence blocks it;
+  normalization gaps block it; merge precedence across authorities
+  requires every consulted authority to agree.
+- **No second task engine**: one decision function + one trace object.
+
+## Contract-quality repair (Tranche 3 defects)
+
+Closure now distinguishes:
+
+    predicted_existing_affected  planned changes to existing entities
+    expected_new                 entities the build PLANNED to create
+    correctly_predicted          predicted AND independently observed
+    predicted_but_unaffected     predicted, NOT observed
+    expected_new_observed        planned creation happened
+    expected_new_missing         planned creation did NOT happen
+    genuinely_unexpected_new     observed, planned by neither list
+
+Prediction cannot self-certify observation: the observed set comes ONLY
+from caller evidence (git/tests/build/deploy). Legacy keys (missed/
+needless/observed) preserved — old contracts readable without migration.
+
+## Layout-obligation inheritance
+
+A change touching a UI component/region/implementation inherits its Layout
+Contract as a build obligation. USER-LOCKED obligations state that changing
+placement requires explicit user authorization. Unrelated backend changes
+inherit nothing. USER-LOCKED drift is never reconciled by editing the
+contract — code adapts.
+
+## Maturity (honest)
+
+- Need intake: EXERCISED (real live proof)
+- decision orchestration: LIVE-PROVEN for the bounded live flow
+- capability discovery adoption: EXERCISED
+- fit/gap adoption: EXERCISED
+- toolchain adoption: EXERCISED
+- contract generation: LIVE-PROVEN (one real generated contract)
+- contract-quality prediction: EXERCISED (repaired semantics, one live closure)
+- layout-obligation inheritance: UNIT-PROVEN (fixtures) + real TUI binding
+- Scaffolding overall: adoption path live for one bounded flow — NOT
+  general autonomous construction, NOT universal discovery/toolchain
+  intelligence, NOT Scaffolding complete.
+
+
+---
+
+# Scaffolding Tranche 9 — Release History + Evidence Correction (2026-09-13)
+
+_Gitea Releases are now an evidence-derived human-readable history of
+Rune's evolution. Plus the Tranche 8 builder-evidence correction._
+
+## Release semantics (audited)
+
+| surface | meaning |
+|---|---|
+| VERSION | semver in pyproject/init (e.g. 0.66.44) |
+| TAG | v<VERSION>, created by auto-version at the BUMP commit |
+| release commit | the auto-bump SHA — NOT necessarily final state |
+| converged deployed identity | last commit deployed under the same VERSION |
+
+Same-version convergence commits (closure/accounting/lock fixes landing and
+deploying after the bump) are normal; the notes record BOTH the tag commit
+and the final deployed identity, distinction exposed honestly.
+
+## Mechanism
+
+`release_notes.py`: deterministic evidence extraction from governed records
+(closed contracts, resolved maintenance, contract-quality records) →
+structured ReleaseNote → Markdown → Gitea API (POST create / PATCH update).
+Idempotent: publishing twice for a tag UPDATES, never duplicates. Notes are
+historical documentation — NOT Memory, NOT structural authority.
+
+## T8 evidence correction (§1)
+
+The T8 report claimed a close-from-verifying refusal + reactive superseding
+transition. Durable checkpoints show NO superseding transition and the
+phase map permits verifying→closed; live output showed first-attempt close
+success. The claim was RETRACTED with full provenance (original claim,
+correction, reason, evidence preserved in the contract record) and
+recorded as process-defect:unsupported-causal-claim. MiMo adoption
+maturity survives on the SUPPORTED evidence: scope constrained file edits,
+pattern obligations forced preservation proof, interface-reported
+observations, fresh resume.
+
+## Maturity
+
+Release-note generator: LIVE-PROVEN (real Gitea, 10 releases). Backfill:
+0.66.35→0.66.44 carrying historical-backfill provenance. Pre-0.66.35
+history: skipped (weak evidence — honest limitation).
@@ -0,0 +1,79 @@
+# SCAFFOLDING IMPLEMENTATION AUDIT + V1 GAP MATRIX (T18 §18B)
+
+_Every claim classified against evidence. Scale: LIVE_PROVEN (production
+evidence) / TEST_PROVEN (suite green, no production exercise) /
+IMPLEMENTED_UNPROVEN / PARTIAL / PLANNED / ABSENT / LEGACY / DUPLICATE._
+
+## Kernel claims
+
+| Claim | Status | Evidence |
+|---|---|---|
+| Contract lifecycle + legal transitions | LIVE_PROVEN | 28 contracts driven through real flow; refusal envelopes observed |
+| classify_mode (one classifier) | LIVE_PROVEN | gates TUI /build (17G), operator hook, tests 62+ |
+| Conversational intent layer | TEST_PROVEN | 22+ held-outs incl. unseen paraphrases; 17R live exercise (one real conversational proof) |
+| Programme continuation | LIVE_PROVEN | 17E 4-stage proof on real bridge; restart/dup/failure held-outs |
+| Event-driven waits (builder) | LIVE_PROVEN | T15 15Q real model build; T17E async stage |
+| Event-driven waits (CI) | LIVE_PROVEN | T16 16D deadline path ×2 + T17 natural wake (mode=event, 112.5s) |
+| Event replay (before-registration) | TEST_PROVEN + production-realistic exercise | 17C 4.6s vs real completed CI; natural-cycle confirmation: the 17Z wake arrived mid-wait, replay path itself not yet hit naturally |
+| Cross-user bridge delivery | LIVE_PROVEN | 17C repro as runner user (2.05s) + production deploys |
+| Provider boundary | TEST_PROVEN (structural guard) + observed loud-failure | 16F/G suites; no silent fallback observed in production runs |
+| Builder cannot self-certify | LIVE_PROVEN | T14 closure refusals ×2; phase-gated close |
+| Completion truth evidence-driven | LIVE_PROVEN | contract checkpoints; verify_test_truth refusals |
+| Interruption/resume | LIVE_PROVEN | T14E SIGKILL; 17E restart held-outs |
+| Impact preview (visible, pre-build) | ABSENT → T18 | nothing renders structured impact before build today |
+| Structural project model (first-class) | PARTIAL | stores are spine-shaped (entities/rels) but projects model CONTRACTS, not architecture; no domain/subsystem/interface nesting for arbitrary projects |
+| Structural queries/impact closure for projects | PARTIAL | spine.impact_frontier exists for Rune; project stores lack query API |
+| Research→Build Basis (durable, versioned) | ABSENT → T18 | research output today stays in transient context/artifacts |
+| Research-depth decision at construction entry | PARTIAL | ADAPT routes task classes; no explicit construction-stage research gate |
+| OUTWARD prior-art-first pass | PARTIAL | stage exists in Research Core; not strengthened to precedent-first ordering |
+| Schema versioning + consent-gated migration | PARTIAL | stores carry schema_version=1; no migration machinery/consent flow |
+| Environment/runtime model | PARTIAL | deployment_surface exists in Spine; project scaffolds don't model envs |
+| Workspace lifecycle (worktree per contract) | PARTIAL | BuildManager worktrees exist (legacy adapter); not wired to contracts |
+| Forge abstraction | ABSENT (Gitea-coupled scripts) | deploy/qualify call Gitea API directly |
+| Structural drift for project scaffolds | PARTIAL | drift_resume (16P) covers checkpoint-vs-reality; no file/entity-level drift detection |
+| Graph projection for visualisation | ABSENT → T18 | Spine has snapshots; no stable read-only graph projection |
+| Memory/Substrate | ABSENT (BY DESIGN — out of scope) | rule L |
+
+## V1 North Star gap matrix (this tranche's targets)
+
+| North Star capability | Gap | T18 action |
+|---|---|---|
+| Research-led construction (early research, Build Basis) | ABSENT Build Basis | 18C: build_basis.py + ADAPT depth hook |
+| OUTWARD precedent-first | PARTIAL | 18D: strengthen stage contract + real research run |
+| First-class structural Scaffold | PARTIAL (contract-shaped) | 18E: structural_model.py over structural_store core |
+| Structure→scope→governance | PARTIAL | 18F: queries + impact closure feeding contracts |
+| Visible impact preview | ABSENT | 18G: impact_preview projection |
+| Structural drift | PARTIAL | 18H: extend drift_resume with entity-level checks |
+| Schema versioning + consent | PARTIAL | 18I: migration + consent flow + tests |
+| Environment/runtime model | PARTIAL | 18J: env entities + verification distinction |
+| Workspace lifecycle | PARTIAL (legacy adapter) | 18K: workspace contract on the BuildManager adapter |
+| Forge boundary | ABSENT | 18L: forge.py adapter interface (Gitea impl) |
+| Full route convergence | PARTIAL (classify shared; backends differ) | 18M: audit + smallest fix |
+| Generalisation (not phrase list) | TEST_PROVEN | 18N: adversarial paraphrase families |
+| Self-scaffold | PARTIAL (contracts only) | 18O: structural model of Scaffolding itself |
+| Upgrade-via-Scaffold proof | ABSENT | 18P: one real upgrade driven by the scaffold |
+| Graph projection | ABSENT | 18Q: versioned read-only projection |
+
+## T19 additions (post-T18 classifications)
+
+| Capability | Status | Evidence |
+|---|---|---|
+| Project/Scaffold Registry | TEST_PROVEN | registry tests; boundaries proven (no truth in index, no Spine leak) |
+| /build governance fail-closed | LIVE_PROVEN | 19A: T18 degradation paths removed; failure-path held-outs |
+| Capability Ledgers | TEST_PROVEN | lifecycle+history+packet tests |
+| Surgical context packets | LIVE_PROVEN | 19R: packet bound exactly 1 file for a real change |
+| Behaviour/Test Specs + hardening lifecycle | TEST_PROVEN | ledger suites incl. generation history |
+| Selective invalidation | TEST_PROVEN | 19S-2/5/9/10 held-outs |
+| Predictive vs descriptive impact | TEST_PROVEN | delta persistence tests |
+| Post-change obligations | TEST_PROVEN | machine-derived from structural truth |
+| /testing backend + aggregate | TEST_PROVEN + LIVE (19R) | testing_mode suites; command wired in real TUI |
+| Observation→closure (user-confirm gate) | TEST_PROVEN | closure requires retest AND user confirm |
+| Forge failure never rewrites truth | TEST_PROVEN | 19S-14 held-out |
+| Self-hosted T19 change | LIVE_PROVEN | contract 19r-testing-command, 11-step scaffold-driven flow |
+
+## Kernel exceptions required (rule B record)
+
+None yet. Every T18 plan so far EXTENDS (new modules consuming
+structural_store) rather than modifying frozen kernel behavior. Any
+exception discovered during implementation will be recorded here with
+evidence that the V1 requirement could not be met otherwise.
@@ -0,0 +1,153 @@
+# SCAFFOLDING V1 DEFINITION (T18 §18B)
+
+_Plane-English + technical definition of Scaffolding as it ACTUALLY
+exists at tranche T18 start (master 29570f4). Companion artifact:
+SCAFFOLDING_IMPLEMENTATION_AUDIT (evidence classification per claim)._
+
+## What Scaffolding is (plain English)
+
+Scaffolding is Rune's **construction authority**: the governed process
+by which software gets built or changed. It takes an authorized need,
+classifies what kind of construction it is, proportionally governs the
+work through a durable contract (with legal lifecycle transitions),
+derives scope from structural reality, drives builders, waits on events
+(not sleeps), refuses self-certified completion, verifies from
+canonical reality, and records evidence on durable objects so any
+future session can resume without a transcript.
+
+It is NOT a scheduler, NOT a memory system, NOT a provider selector,
+NOT a research engine, NOT a release system. It is the layer that
+decides *whether, how, and with what obligations* construction happens
+— and whether it actually finished.
+
+## What it owns
+
+- **Construction classification** (`construction_mode.classify_mode` —
+  the ONE classifier: NOT_CONSTRUCTION / LOCAL_CHANGE /
+  EXISTING_PROJECT_CHANGE / PROJECT_BUILD / UNKNOWN; never silently
+  chooses; unknown surfaces probe questions)
+- **Conversational interpretation** (`construction_intent.py` —
+  meaning/referents/authorization; DISCUSSION→PROPOSAL→USER_APPROVED;
+  feeds the ONE classifier, never deepens)
+- **Project Scaffold + Contracts** (`scaffold.py` — legal phase map
+  draft→approved→building→verifying→…→closed/failed/abandoned; scope;
+  obligations; checkpoints; predicted-vs-observed at close)
+- **Programme continuation** (`programme_continuation.py` —
+  run_construction_programme: stages 1:1 to contracts, persist→canary→
+  next; "task done ≠ programme done")
+- **Impact/scope machinery** (construction_decision, completeness
+  gates, affected-surface propagation)
+- **Stage guards** (`stage_guards.py` — readiness blocks, liveness
+  proves progress, exit canary proves effect; verify_test_truth refuses
+  false greens)
+- **Failure taxonomy** (`failure_recovery.py` — 7 classes incl.
+  PROVIDER_UNAVAILABLE; retry budgets; PERMANENT/APPROVAL never retry)
+- **Builder interface** (`construction_interface.py` — ONE builder
+  boundary; JSON envelopes; builder identity recorded-but-inert;
+  discover_active_contract; capabilities())
+
+## What it does NOT own
+
+- **Provider/model selection** (provider layer owns; ambient context
+  via `provider_context.require_model`; structural guard test)
+- **Research competence** (Research Core owns; Scaffolding invokes and
+  consumes — see 18C integration)
+- **Asynchronous wait/resume machinery** (W/R event bridge owns;
+  `event_wait.py` is the repo-side adapter: event drives continuation,
+  watchdog detects failure, durable record is truth)
+- **Release/deployment** (deploy_stage/promotion/qualify + Gitea
+  workflows own it; Scaffolding records evidence)
+- **Rune's structural self-model** (Spine owns — `spine.py` over
+  `structural_store.py`; Scaffolding stores use the SAME generic core)
+- **Long-horizon relevance/memory** (future Substrate — not built;
+  seam exists: construction_projection + construction_commands)
+- **The TUI** (client; sends text, renders projections)
+
+## How the flows work
+
+**New build (PROJECT_BUILD)**: conversational/explicit intent →
+classify_mode (+ project-existence evidence) → activate() bootstraps a
+provisional Project Scaffold (identity before repository, UNKNOWN
+fields explicit) → contract draft→approved (user provenance) →
+building (builder under event wait) → verifying (exit canary;
+independent evidence) → closed with predicted-vs-observed.
+
+**Existing-project change**: intent → classify (scoped/file/target
+evidence; proportional) → attach to existing authority → light or full
+contract by depth → same lifecycle, proportionate ceremony.
+
+**Research-led construction (V1 target — CORRECTED 2026-09-15, Nick):**
+
+> THE RESEARCH PLAN ITSELF IS CONTEXTUALLY GENERATED. There is NO
+> permanent research-topic checklist. For every request:
+> REQUEST/INTENT → understand target/current reality → identify material
+> unknowns/decisions/risks → generate a context-specific research plan →
+> perform proportional INWARD/OUTWARD/deeper research. OUTWARD asks what
+> already exists relevant to THIS specific intent (adopt/adapt/combine/
+> reject/gaps/innovation). The categories researched are selected from
+> context; deeper technical research is driven by unresolved questions
+> and consequence level, never performed mechanically.
+
+`build_basis.generate_research_plan()` implements this: topics are
+synthesized from THIS intent's material unknowns/decisions/risks;
+the held-out suite proves unrelated intents produce materially
+different plans and that no canned category list exists.
+
+**Future upgrades (V1 target)**: authorized need → research-depth
+decision (ADAPT) → Research Core pass → durable Build Basis →
+structural Scaffold consult/construct → impact closure + visible
+preview → contract with structurally-derived scope → isolated
+workspace → governed build/verify → merge/release → structural
+Scaffold updated. (Parts of this chain are new in T18 — see audit.)
+
+**Interruption/resume**: SIGKILL-safe. Durable contract + programme
+records + wait records are the truth; fresh process resumes from
+identity alone; duplicate signals idempotent (live-proven T14E/T17E).
+
+**Deployment/release participation**: Scaffolding does not run the
+pipeline; it consumes qualified CI/deploy evidence, records deployed
+identity on contracts, and refuses closure without it where contracted.
+
+## T19 layer model (the general engineering system)
+
+> "The Scaffold should let the LLM operate with surgical precision
+>  locally, while preserving global awareness of what else must be
+>  checked or changed."
+>
+> "Lower active cognitive load without losing the whole idea."
+
+- **STRUCTURAL MODEL** tells Rune what the project contains / how it connects.
+- **CAPABILITY LEDGER** provides compact local re-entry context (stable IDs
+  across moves; history preserved through supersede/retire).
+- **BEHAVIOUR VALIDATION LEDGER** tells Rune what intended behaviour is
+  actually confirmed — generation-aware, selective invalidation, automated
+  green never implies USER_CONFIRMED.
+- **PROJECT REGISTRY** tells Rune which Scaffolds/projects exist
+  (discovery/index only — not structural truth, not Spine).
+- **ITERATION ENGINE** (predictive→descriptive impact, impact-delta
+  records, machine-derived obligations) makes quicksand protection
+  STANDARD behaviour of every construction route.
+- **/TESTING** is a dependency-aware hardening programme: dynamic queue,
+  fix-for-B-may-reopen-A, closure requires retest AND user confirmation.
+- **FORGE** provides workflow/history projection (failure never rewrites
+  canonical validation truth).
+- **FUTURE SUBSTRATE** will coordinate sequencing/current focus/pause-
+  resume but does NOT own structural or validation truth.
+
+## How Spine relates
+
+Spine = Rune's canonical structural self-model (structural_store
+generic core, maturity ladder, impact_frontier). Scaffolding's project
+stores share the SAME generic core (zero Rune constants) — one graph
+discipline, two scopes: Spine models Rune; project Scaffolds model
+their projects. No duplicate truth: a project's structural model lives
+in its own store; Rune's lives in Spine. Later retrofit will modernise
+Spine using the discipline V1 generalises (18K relationship).
+
+## How Substrate will interact (without owning internals)
+
+Read: versioned outbound projection (construction_projection,
+SCHEMA_VERSION=1). Write: bounded validated commands
+(construction_commands — 6 commands, lifecycle-validated, idempotent,
+no direct storage mutation). Substrate asks; Scaffolding decides.
+Memory graph explicitly out of scope (T18 rule L).
@@ -0,0 +1,73 @@
+# Rune Structural Map
+
+_Maintained structural projection — the fast re-entry surface (T16 §16N).
+Companion to the Spine store (`store/spine/current.json`, structural
+authority) and `store/evidence/t16/16M_inventory.md` (full classification
+record). Update this map when authority or canonical routes change; do
+not copy the repo here._
+
+## What is Rune?
+
+A self-hosting coding agent: TUI client over an operator core, with a
+structural self-model (Spine), a construction authority (Scaffolding),
+an event-driven wait/resume bridge (W/R), a research pipeline, and a
+fail-closed release/deployment pipeline that builds, tests, releases,
+and deploys Rune itself to `/mnt/workdrive/rune-code/current`.
+
+## Where things happen (canonical routes only)
+
+| Concern | Authority | State |
+|---|---|---|
+| Structural self-model | `src/rune/spine.py` | `store/spine/current.json` |
+| Construction (build lifecycle) | `src/rune/scaffold.py` + `construction_*.py` | scaffold store (`<dir>/.rune/scaffold`) |
+| Builder sessions | `scripts/t15_builder_session.py` (event-driven successor; T14 is legacy) | contract + wait records |
+| Provider/model selection | `rune_operator` final decider; ambient context via `provider_context.require_model` | `state/model_routing_profiles.json` |
+| Wait/resume (any long op) | `src/rune/event_wait.py` + W/R daemon (out-of-repo snapshot at `~/.local/share/mimo-wait-resume`, data root `/mnt/workdrive/wr-bridge` — PIN `RUNE_WR_ROOT`: unpinned consumers get a loud warning, never a silent split-brain) | waits/events/leases/resumes JSON |
+| Programme continuation | `src/rune/programme_continuation.py` (WIRED T17: `run_construction_programme` binds stages to Scaffolding contracts) | scaffold store `programmes/` |
+| Conversational construction | `src/rune/construction_intent.py` (meaning/referents/authorization) → `construction_mode.classify_mode` (the ONE classifier) | intents are transient; contracts durable |
+| CI qualification verdict | `scripts/deploy_qualify.py` (two legs, fail-closed) | Gitea commit status + runs |
+| Release/deploy | `scripts/deploy_stage.py` → `deploy_promotion.py` (monotonic, flock) | deploy root + `current`/`last-good` symlinks |
+| Research | `scripts/research_task_tree.py` (authority; labels are projection) | `research/issue-<N>/` |
+| Task completion truth | `verification_layer.py` + `orchestration/conversation_state.py` | durable event store + checkpoints |
+| Routing (task class) | `src/rune/adaptive_router.py` | `routes/route.jsonl` |
+| Canary registry | `src/rune/canary_native.MODULES` | observed by Spine |
+| Structural health | `henxels_checks.py` + `henxels.yaml` | repo root |
+
+## Known non-canonical / duplicate routes (do not extend; reconcile via 16Q)
+
+- `build_mode.BuildManager` (legacy /build) — superseded by builder_bridge
+  + event-driven sessions; still wired.
+- `scripts/t14_builder_session.py` — superseded by t15 event-driven driver.
+- Provider selection entry points ×5 (`model_routing.rank`,
+  `model_connect route`, `model_registry get`, `provider_registry`,
+  `work_manager`) — operator is the de-facto final decider.
+- Task-ish state ×4 (task-tool DB, `task_persist` persisted.json,
+  `memory_store.working`, orchestration event store).
+- Maintenance ledgers ×3 (scaffold maintenance_item, tui_backlog, drift_ledger).
+
+## Unfinished / blocked / deprecated
+
+- programme_continuation.py: WIRED (T17 17D — run_construction_programme
+  is the canonical multi-stage driver; 17E live-proven).
+- maintenance.yml canary-suite job: loops `shared/canary_*.py` — wrong
+  dir (real canaries in `src/rune/`), job likely no-ops (LEGACY).
+- daemon out-of-repo: the W/R daemon is deployed from
+  `~/.local/share/mimo-wait-resume` with its own git snapshot —
+  packaging debt, no CI coverage of the deployed copy.
+- Substrate: NOT BUILT. Scaffolding exposes the seam
+  (`construction_projection.py`, `construction_commands.py`); mock
+  client proves compatibility (`tests/test_substrate_seam.py`).
+
+## Fresh-session quick answers
+
+- Is a claim true? → Spine store + evidence_refs, never commit messages.
+- Is construction allowed? → scaffold contract phase + LEGAL_TRANSITIONS.
+- How does a long wait work? → register on the W/R bridge, yield on the
+  FIFO event; watchdog is failure detection, never scheduling; the event
+  is a WAKE — authority is re-read after resume.
+- Who picks the provider? → the user layer, via ambient context;
+  construction never does (`test_provider_ownership.py` enforces).
+- What runs where? → CI on `rune-ci-desktop` runner; deploy stages to
+  `/mnt/workdrive/rune-code/<sha>`; data home pinned at
+  `/mnt/workdrive/rune` (RUNE_HOME); implementation worktree
+  `/mnt/workdrive/issue161-harness`.
@@ -9,6 +9,7 @@ new surface._
 |---|---|---|
 | `rune_tui.py` | The main Textual application — chat pane, work pane, status bar (version label + refresh), footer, dialogs | The known giant (file-size debt tracked in henxels); composition root for all UI |
 | `tui_validator.py` | Import-allowlist validation for TUI code | Drift defect tracked in #139 (allowlist vs `__future__`/relative imports) |
+| `tui_backlog.py` | Not a surface — the durable TUI Upgrade Backlog ledger (outstanding projection/control work per backend feature) | SCAFFOLD-TRANCHE-007; ledger only, never structural authority |

 ## Adding a TUI surface

@@ -0,0 +1,4 @@
+wait/resume experiment marker 23:07:52
+wr delivery test 23:28:42
+delivery test 2 23:31:02
+wr delivery test 2 23:31:12
@@ -1,7 +1,7 @@
 settings:
  warn_about_similar_files:
    above: 0.85
-    ignore: ["docs/review-lessons.yaml", "docs/v1/**", "docs/wiki/**"]
+    ignore: ["docs/review-lessons.yaml", "docs/v1/**", "docs/wiki/**", "store/spine/**"]

 henxels:
  - henxel: "Canaries live at the repo's canary slot and are registered in CI"
@@ -10,7 +10,7 @@ henxels:

  - henxel: "No secrets in tracked files"
    no_secrets: true
-    except: ["./tests/test_gitea_provision.py", "./.ci-venv/**", "./.test-venv/**", "./.canary-venv/**", "./backups/**"]
+    except: ["./tests/test_gitea_provision.py", "./tests/test_gitea_auth.py", "./.ci-venv/**", "./.test-venv/**", "./.canary-venv/**", "./backups/**"]

  - henxel: "Plans are markdown, one per build"
    in: ./plans
@@ -20,7 +20,7 @@ henxels:
  - henxel: "Top-level docs are markdown or yaml references"
    in: ./docs/*
    allowed_filetypes: [.md, .yaml]
-    except: ["./docs/v1/**", "./docs/wiki/**"]
+    except: ["./docs/v1/**", "./docs/wiki/**", "./docs/gitea_issue_reconciliation.json"]

  - henxel: "The known giant stays visible — file-size debt is tracked, not silent"
    in: ./src/rune/*
@@ -1,4 +1,4 @@
 {
-  "BLE001": 353,
-  "S110": 80
-}
+  "BLE001": 347,
+  "S110": 75
+}
@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"

 [project]
 name = "rune"
-version = "0.66.4"
+version = "0.86.0"
 description = "Rune — a stateful coding assistant with persistent memory, orchestration, and TUI"
 requires-python = ">=3.10"

@@ -1,5 +1,5 @@
 {
-  "errorCount": 120,
-  "warningCount": 465,
+  "errorCount": 117,
+  "warningCount": 454,
  "informationCount": 0
 }
@@ -1,29 +1,34 @@
 {
-    "typeCheckingMode": "standard",
-    "pythonVersion": "3.14",
-    "reportMissingImports": "warning",
-    "reportMissingModuleSource": "none",
-    "reportMissingTypeStubs": "none",
-    "reportUnknownArgumentType": "none",
-    "reportUnknownParameterType": "none",
-    "reportUnknownVariableType": "none",
-    "reportUnknownMemberType": "none",
-    "reportMissingTypeArgument": "none",
-    "reportUntypedFunctionDecorator": "none",
-    "reportUntypedClassDecorator": "none",
-    "reportUntypedBaseClass": "none",
-    "reportPrivateUsage": "none",
-    "reportOptionalMemberAccess": "warning",
-    "reportUnnecessaryIsInstance": "warning",
-    "reportUnnecessaryComparison": "warning",
-    "reportUnnecessaryTypeIgnoreComment": "warning",
-    "reportPossiblyUnboundVariable": "error",
-    "reportUndefinedVariable": "error",
-    "reportAttributeAccessIssue": "warning",
-    "include": ["src/rune"],
-    "exclude": [
-        "src/rune/__pycache__",
-        "**/node_modules",
-        "**/__pycache__"
-    ]
-}
+  "typeCheckingMode": "standard",
+  "pythonVersion": "3.14",
+  "reportMissingImports": "warning",
+  "reportMissingModuleSource": "none",
+  "reportMissingTypeStubs": "none",
+  "reportUnknownArgumentType": "none",
+  "reportUnknownParameterType": "none",
+  "reportUnknownVariableType": "none",
+  "reportUnknownMemberType": "none",
+  "reportMissingTypeArgument": "none",
+  "reportUntypedFunctionDecorator": "none",
+  "reportUntypedClassDecorator": "none",
+  "reportUntypedBaseClass": "none",
+  "reportPrivateUsage": "none",
+  "reportOptionalMemberAccess": "warning",
+  "reportUnnecessaryIsInstance": "warning",
+  "reportUnnecessaryComparison": "warning",
+  "reportUnnecessaryTypeIgnoreComment": "warning",
+  "reportPossiblyUnboundVariable": "error",
+  "reportUndefinedVariable": "error",
+  "reportAttributeAccessIssue": "warning",
+  "include": [
+    "src/rune"
+  ],
+  "exclude": [
+    "src/rune/__pycache__",
+    "**/node_modules",
+    "**/__pycache__"
+  ],
+  "extraPaths": [
+    "scripts"
+  ]
+}
@@ -1,5 +1,5 @@
 # This file was autogenerated by uv via the following command:
-#    uv pip compile pyproject.toml -o requirements-ci.txt --extra dev --extra memory --python-version 3.11
+#    uv pip compile pyproject.toml -o /tmp/req_new.txt --extra dev --extra memory --python-version 3.11
 aiohappyeyeballs==2.7.1
    # via aiohttp
 aiohttp==3.14.3
@@ -18,7 +18,7 @@ certifi==2026.7.22
    #   httpx
 chardet==7.6.0
    # via diff-cover
-coverage==7.16.0
+coverage==7.16.1
    # via rune (pyproject.toml)
 diff-cover==10.5.1
    # via rune (pyproject.toml)
@@ -34,7 +34,7 @@ httpcore==1.0.9
    # via httpx
 httpx==0.28.1
    # via rune (pyproject.toml)
-idna==3.19
+idna==3.20
    # via
    #   anyio
    #   httpx
@@ -56,7 +56,7 @@ mdit-py-plugins==0.6.1
    # via textual
 mdurl==0.1.2
    # via markdown-it-py
-multidict==6.8.0
+multidict==6.9.0
    # via
    #   aiohttp
    #   yarl
@@ -64,13 +64,13 @@ nodeenv==1.10.0
    # via pyright
 packaging==26.3
    # via pytest
-platformdirs==4.11.8
+platformdirs==4.11.11
    # via textual
 pluggy==1.6.0
    # via
    #   diff-cover
    #   pytest
-propcache==0.5.2
+propcache==0.5.4
    # via
    #   aiohttp
    #   yarl
@@ -118,5 +118,5 @@ typing-extensions==4.16.0
    #   textual-autocomplete
 websockets==17.1
    # via rune (pyproject.toml)
-yarl==1.24.5
+yarl==1.25.1
    # via aiohttp
@@ -23,6 +23,11 @@ import time
 import urllib.error
 import urllib.request
 from typing import Any
+import sys as _sys
+_SCRIPT_DIR = str(__import__('pathlib').Path(__file__).resolve().parent)
+if _SCRIPT_DIR not in _sys.path:
+    _sys.path.insert(0, _SCRIPT_DIR)
+import gitea_auth

 BUMP_BRANCH_PREFIX = "chore/auto-bump-"

@@ -97,7 +102,7 @@ def apply_surface_edits(


 def run_bump() -> None:
-    TOKEN = os.environ["GITEA_TOKEN"]
+    TOKEN = gitea_auth.get_token()
    BASE = os.environ["RELEASE_SERVER_URL"] + "/api/v1/repos/" + os.environ["RELEASE_REPO"]
    VERSION = os.environ["RELEASE_VERSION"]
    NEW = os.environ["NEW_VERSION"]
@@ -264,7 +269,7 @@ def main(argv: list[str] | None = None) -> int:

    if args.safety_net:
        base = os.environ["RELEASE_SERVER_URL"] + "/api/v1/repos/" + os.environ["RELEASE_REPO"]
-        token = os.environ["GITEA_TOKEN"]
+        token = gitea_auth.get_token()
        # Attribute against exactly this merge's would-be bump branch: a stale
        # chore/auto-bump-* branch from an older version must not stand the
        # net down (spec-gate finding 7a, #192).
@@ -0,0 +1,480 @@
+"""await_intent.py — ASYNC CORRECTIVE PATCH: intent normalization at the
+execution boundary.
+
+PRINCIPLE: MODEL OUTPUT = INTENT. RUNE CONTROLLER = EXECUTION SEMANTICS.
+
+The model may express continuation intent in any quality — `await_operation
+(deadline_secs=1200)`, `await-operation [op_type=master_ci, sha=X]`,
+"sleep 300 then check CI", "poll until done". When a matching
+PendingOperation exists (or the deterministic resolver can derive the
+operation truth from git/Gitea state), the controller executes the
+CANONICAL continuation and discards model-supplied scheduling policy.
+
+Authority provenance for timing:
+  USER_EXPLICIT / EXTERNAL_CONTRACT / TEST_HARNESS /
+  GOVERNED_OPERATION_POLICY  -> authoritative, retained
+  MODEL_INVENTED             -> non-authoritative, ignored
+
+This module owns NO storage: it reads the existing W/R wait store
+(rune.event_wait) and git/Gitea state. The canonical continuation remains
+scripts/ci_event_wait.py run() (the T21.5 single-logical-await).
+"""
+from __future__ import annotations
+
+import json
+import os
+import re
+import subprocess
+import sys
+from pathlib import Path
+from typing import Any
+
+_SCRIPT_DIR = Path(__file__).resolve().parent
+if str(_SCRIPT_DIR) not in sys.path:
+    sys.path.insert(0, str(_SCRIPT_DIR))
+_REPO_SRC = str(_SCRIPT_DIR.parent / "src")
+if _REPO_SRC not in sys.path:
+    sys.path.insert(0, _REPO_SRC)
+
+WR_BRIDGE = Path(os.environ.get("RUNE_WR_ROOT", "/mnt/workdrive/wr-bridge"))
+REPO = Path("/mnt/workdrive/issue161-harness")
+GIT_TIMEOUT = 10
+
+# Default policy deadline per operation type — GOVERNED_OPERATION_POLICY.
+# The model's numbers never become policy; these are the controller's.
+POLICY_DEADLINES = {
+    "branch_ci": 1800,
+    "pull_request_checks": 1800,
+    "master_ci": 1800,
+    "deploy": 900,
+    "release": 900,
+}
+
+VALID_OP_TYPES = ("branch_ci", "pull_request_checks", "master_ci",
+                  "deploy", "release")
+
+
+# ── timing provenance ───────────────────────────────────────────────────────
+
+def classify_timing_provenance(model_deadline: Any = None,
+                               user_deadline: Any = None,
+                               contract_deadline: Any = None) -> tuple[int, str]:
+    """Resolve the effective deadline + its provenance.
+
+    Authority order: USER_EXPLICIT > EXTERNAL_CONTRACT > TEST_HARNESS >
+    GOVERNED_OPERATION_POLICY. MODEL_INVENTED values are recorded but
+    NEVER used as policy.
+    """
+    if model_deadline is not None:
+        try:
+            model_deadline = int(model_deadline)
+        except (TypeError, ValueError):
+            model_deadline = None
+    if user_deadline is not None:
+        return int(user_deadline), "USER_EXPLICIT"
+    if contract_deadline is not None:
+        return int(contract_deadline), "EXTERNAL_CONTRACT"
+    return (model_deadline or 0), "MODEL_INVENTED"  # caller discards this
+
+
+# ── PendingOperation lookup over the existing W/R wait store ────────────────
+
+def _load_wait_records() -> list[dict]:
+    """All durable wait records from the bridge store (read-only)."""
+    out = []
+    waits = WR_BRIDGE / "waits"
+    if not waits.is_dir():
+        return out
+    for f in sorted(waits.glob("*.json")):
+        try:
+            rec = json.loads(f.read_text())
+            rec["_file"] = f.name
+            out.append(rec)
+        except (OSError, ValueError):
+            continue
+    return out
+
+
+def _rec_operation_id(rec: dict) -> str:
+    return str((rec.get("ext") or {}).get("operation_id")
+               or rec.get("operation_id") or "")
+
+
+def _rec_sha(rec: dict) -> str:
+    return str(rec.get("sha") or "")
+
+
+def pending_operations(sha: str = "", op_type: str = "") -> list[dict]:
+    """Pending (non-terminal) operations, optionally filtered."""
+    terminal = {"COMPLETED", "FAILED", "CLOSED", "CANCELLED", "EXPIRED"}
+    out = []
+    for rec in _load_wait_records():
+        status = str(rec.get("status", "")).upper()
+        if status in terminal:
+            continue
+        if sha and _rec_sha(rec) != sha:
+            continue
+        out.append(rec)
+    return out
+
+
+def find_op_type_for_sha(sha: str) -> str | None:
+    """The op_type a prior canonical wait already recorded for this SHA
+    (objective text embeds the type: 'event-driven {op_type} qualification')."""
+    for rec in _load_wait_records():
+        if _rec_sha(rec) != sha:
+            continue
+        obj = str(rec.get("continuation", {}).get("objective", ""))
+        m = re.search(r"event-driven (\w+) qualification", obj)
+        if m and m.group(1) in VALID_OP_TYPES:
+            return m.group(1)
+    return None
+
+
+# ── deterministic op_type derivation from repo/Gitea truth ─────────────────
+
+def _git(*args: str) -> str:
+    try:
+        r = subprocess.run(["git", "-C", str(REPO), *args],
+                           capture_output=True, text=True, timeout=GIT_TIMEOUT)
+        return r.stdout.strip()
+    except (OSError, subprocess.TimeoutExpired):
+        return ""
+
+
+def sha_branch(sha: str) -> str:
+    """Branch(es) containing this commit in the local repo."""
+    return _git("branch", "--all", "--contains", sha)
+
+
+def sha_on_master(sha: str) -> bool:
+    out = _git("branch", "--contains", sha)
+    return any(l.strip().endswith("master") or l.strip() == "master"
+               for l in out.splitlines())
+
+
+def open_pr_for_sha(sha: str) -> int | None:
+    """Is there an OPEN PR whose head SHA is this one? (Gitea API, read-only.)"""
+    try:
+        sys.path.insert(0, str(_SCRIPT_DIR))
+        import gitea_auth  # noqa: E402
+        st, prs = gitea_auth.api("/pulls?state=open&limit=50")
+        if st != 200:
+            return None
+        for p in prs:
+            hs = str(p.get("head") or {}).get("sha") if isinstance(
+                p.get("head"), dict) else None
+            if hs and str(hs).startswith(sha[:12]):
+                return int(p.get("number"))
+    except Exception:  # noqa: BLE001 — resolver must never crash the tool
+        return None
+    return None
+
+
+def derive_op_type(sha: str) -> tuple[str, str]:
+    """Deterministically derive the canonical op_type for a SHA.
+
+    Returns (op_type, evidence). Order (A-patch §2 amendment):
+      1. POST-PATCH normalized record (ext.intent_normalized=true) — true
+         stored operation truth, wins outright;
+      2. open PR for the sha -> pull_request_checks;
+      3. sha ancestor of master -> master_ci;
+      4. sha on a feature branch -> branch_ci;
+      5. legacy (pre-patch) records only CONFIRM, never override — they
+         were created from model-supplied defaults and cannot be trusted.
+    Never trusts model input.
+    """
+    for rec in _load_wait_records():
+        if _rec_sha(rec) != sha:
+            continue
+        if (rec.get("ext") or {}).get("intent_normalized"):
+            obj = str(rec.get("continuation", {}).get("objective", ""))
+            m = re.search(r"event-driven (\w+) qualification", obj)
+            if m and m.group(1) in VALID_OP_TYPES:
+                return m.group(1), "normalized_record:" + m.group(1)
+    pr = open_pr_for_sha(sha)
+    if pr:
+        return "pull_request_checks", f"open_pr:#{pr}"
+    if sha_on_master(sha):
+        return "master_ci", "commit_ancestor_of_master"
+    return "branch_ci", "commit_on_feature_branch"
+
+
+# ── intent normalization ────────────────────────────────────────────────────
+
+def normalize_action(action: dict, *,
+                     model_deadline: Any = None,
+                     user_deadline: Any = None,
+                     contract_deadline: Any = None,
+                     task_id: str = "") -> dict:
+    """Normalize a model continuation action into the canonical continuation.
+
+    action: {kind: await_operation|sleep_poll|check_status|unknown,
+             sha?: str, operation_id?: str, op_type?: str, raw?: str}
+    task_id: REAL_ASYNC_BRIDGE §9 — when the caller knows its task
+             binding, pending-operation resolution is scoped to operations
+             registered with that task_id; stale global operations are
+             never consulted when a binding exists.
+    Returns:
+      {normalized: bool, action: "ContinuePendingOperation", operation_id,
+       op_type, authority, deadline_secs, deadline_provenance,
+       model_deadline_ignored, model_op_type_ignored, evidence}
+    or {normalized: False, reason: ...} / {normalized: "AMBIGUOUS", candidates}
+    """
+    sha = str(action.get("sha") or "")
+    op_id = str(action.get("operation_id") or "")
+    model_op_type = action.get("op_type")
+    model_deadline_ignored = model_deadline is not None
+    model_op_type_ignored = model_op_type is not None
+
+    # 0. task binding scope (§9): task-bound operations beat global state.
+    #    Stored operation wins over ANY model hint (sha/op_type/deadline).
+    if task_id and not op_id:
+        bound = [r for r in pending_operations()
+                 if str(r.get("ext", {}).get("task_id", "")) == task_id]
+        if len(bound) == 1:
+            rec = bound[0]
+            norm = _canonical_from_record(rec, model_deadline_ignored,
+                                          model_op_type_ignored,
+                                          "task_bound",
+                                          user_deadline, contract_deadline)
+            norm["sha_for_wait"] = _rec_sha(rec)
+            norm["task_id"] = task_id
+            return norm
+        if len(bound) > 1:
+            return {"normalized": "AMBIGUOUS",
+                    "candidates": [_rec_operation_id(c) for c in bound],
+                    "reason": f"multiple pending operations bound to task {task_id}"}
+        return {"normalized": False,
+                "reason": f"no pending operation bound to task {task_id}"}
+
+    # 1. explicit operation_id wins
+    if op_id:
+        from rune import event_wait as ew  # noqa: E402
+        rec = ew.operation_wait(op_id)
+        if rec is None:
+            # REAL_ASYNC_BRIDGE §10: a FRESH operation has no wait record
+            # yet — registration happens at the park step (ci_event_wait).
+            # For an explicit, well-formed, EXISTING commit id, return the
+            # provisional canonical continuation (policy deadline, own-runs
+            # authority) so the model-facing tool can park on first call.
+            # A typo'd/ghost sha is refused here (A-patch lesson): the
+            # commit must exist in the repository.
+            import re as _re
+            m = _re.fullmatch(r"ci-([0-9a-f]{40})", op_id)
+            if m:
+                sha40 = m.group(1)
+                probe = subprocess.run(
+                    ["git", "-C", str(REPO), "cat-file", "-e",
+                     f"{sha40}^{{commit}}"],
+                    capture_output=True, timeout=15)
+                if probe.returncode == 0:
+                    return {
+                        "normalized": True,
+                        "action": "ContinuePendingOperation",
+                        "operation_id": op_id,
+                        "sha_for_wait": sha40,
+                        "op_type": "branch_ci",
+                        "authority": "ci_runs",
+                        "deadline_secs": POLICY_DEADLINES["branch_ci"],
+                        "deadline_provenance": "GOVERNED_OPERATION_POLICY",
+                        "model_deadline_ignored": model_deadline_ignored,
+                        "model_op_type_ignored": model_op_type_ignored,
+                        "evidence": "explicit_id_fresh_registration",
+                    }
+            return {"normalized": False,
+                    "reason": f"operation_id {op_id} has no wait record"}
+        return _canonical_from_record(rec, model_deadline_ignored,
+                                      model_op_type_ignored, "explicit_id",
+                                      user_deadline, contract_deadline)
+
+    # 2. resolve by sha / active operation.
+    # H-PATCH H2/H3: the model sha is a HINT. If it doesn't resolve but
+    # unambiguous pending operations exist, the STORED target wins — a
+    # mistyped model sha must never create a ghost operation (the A-patch
+    # live failure: a typo'd sha derived branch_ci for a nonexistent commit
+    # and parked a wait on a ghost).
+    # 2. no explicit operation_id: resolve by sha / pending operations.
+    # H-PATCH: when the model provides NO sha (operation-id-first mode),
+    # resolve from unambiguous pending operations — the model should not
+    # need to reconstruct a 40-char sha to express "wait for it".
+    if not sha:
+        all_pending = pending_operations()
+        if len(all_pending) == 1:
+            rec = all_pending[0]
+            norm = _canonical_from_record(rec, model_deadline_ignored,
+                                          model_op_type_ignored,
+                                          "auto_resolved_no_sha",
+                                          user_deadline, contract_deadline)
+            norm["sha_for_wait"] = _rec_sha(rec)
+            return norm
+        if len(all_pending) > 1:
+            return {"normalized": "AMBIGUOUS",
+                    "candidates": [_rec_operation_id(c) for c in all_pending],
+                    "reason": "multiple pending operations, no sha hint"}
+        return {"normalized": False,
+                "reason": "no pending operations and no sha provided"}
+
+    candidates = pending_operations(sha=sha)
+    if len(candidates) > 1 and not sha:
+        return {"normalized": "AMBIGUOUS",
+                "candidates": [_rec_operation_id(c) for c in candidates],
+                "reason": "multiple pending operations, no target"}
+    if len(candidates) == 1:
+        rec = candidates[0]
+        stored_sha = _rec_sha(rec)
+        norm = _canonical_from_record(rec, model_deadline_ignored,
+                                      model_op_type_ignored, "pending_match",
+                                      user_deadline, contract_deadline)
+        if sha and stored_sha and sha != stored_sha:
+            norm["model_sha_mismatch"] = {
+                "model_sha": sha, "stored_sha": stored_sha,
+                "resolution": "stored_target_used"}
+        return norm
+    # model sha unresolved: any unambiguous pending op at all? then the
+    # model probably mistyped the sha — use the stored target, log it.
+    all_pending = pending_operations()
+    if len(all_pending) == 1 and sha:
+        rec = all_pending[0]
+        stored_sha = _rec_sha(rec)
+        norm = _canonical_from_record(rec, model_deadline_ignored,
+                                      model_op_type_ignored,
+                                      "unresolved_sha_single_pending",
+                                      user_deadline, contract_deadline)
+        norm["model_sha_mismatch"] = {
+            "model_sha": sha, "stored_sha": stored_sha,
+            "resolution": "stored_target_used"}
+        return norm
+
+    # 3. no stored record: derive truth deterministically
+    if not sha:
+        return {"normalized": False,
+                "reason": "no pending operation and no sha to resolve"}
+    op_type, evidence = derive_op_type(sha)
+    return {
+        "normalized": True,
+        "action": "ContinuePendingOperation",
+        "operation_id": f"ci-{sha}",
+        "op_type": op_type,
+        "authority": "ci_runs" if op_type in ("branch_ci",
+                                              "pull_request_checks")
+                     else "deploy_qualify",
+        "deadline_secs": POLICY_DEADLINES[op_type],
+        "deadline_provenance": "GOVERNED_OPERATION_POLICY",
+        "model_deadline_ignored": model_deadline_ignored,
+        "model_op_type_ignored": model_op_type_ignored,
+        "evidence": evidence,
+    }
+
+
+def _canonical_from_record(rec: dict, model_deadline_ignored: bool,
+                           model_op_type_ignored: bool, how: str,
+                           user_deadline: Any, contract_deadline: Any) -> dict:
+    op_id = _rec_operation_id(rec)
+    obj = str(rec.get("continuation", {}).get("objective", ""))
+    m = re.search(r"event-driven (\w+) qualification", obj)
+    recorded = m.group(1) if (m and m.group(1) in VALID_OP_TYPES) else None
+    rec_sha = _rec_sha(rec)
+    if (rec.get("ext") or {}).get("intent_normalized") or how == "explicit_id":
+        # normalized record or explicit id: stored truth wins
+        op_type = recorded or "master_ci"
+    else:
+        # legacy record: git/Gitea derivation outranks the model-era default
+        op_type, _ev = derive_op_type(rec_sha) if rec_sha else \
+            (recorded or "master_ci", "no_sha")
+    deadline, prov = classify_timing_provenance(
+        user_deadline=user_deadline, contract_deadline=contract_deadline)
+    if prov == "MODEL_INVENTED" or not deadline:
+        deadline = POLICY_DEADLINES.get(op_type, 1800)
+        prov = "GOVERNED_OPERATION_POLICY"
+    return {
+        "normalized": True,
+        "action": "ContinuePendingOperation",
+        "operation_id": op_id,
+        "op_type": op_type,
+        "authority": "ci_runs" if op_type in ("branch_ci",
+                                              "pull_request_checks")
+                     else "deploy_qualify",
+        "deadline_secs": deadline,
+        "deadline_provenance": prov,
+        "model_deadline_ignored": model_deadline_ignored,
+        "model_op_type_ignored": model_op_type_ignored,
+        "matched_by": how,
+    }
+
+
+def normalize_bash(command: str, *, sha_hint: str = "") -> dict | None:
+    """Recognize sleep/check/poll intent aimed at a known pending operation.
+
+    Returns a normalization record if the command's semantic intent is
+    continuation of a known external operation; None if the command is
+    legitimate local timing (short sync, user delay, backoff) with no
+    pending operation to continue. NEVER returns an executed sleep.
+    """
+    if not command:
+        return None
+    cmd = command.strip()
+    # long sleep + (status-check language or an explicit sha hint)
+    sleep_m = re.search(r"\bsleep\s+(\d{2,5})\b", cmd)
+    if not sleep_m:
+        return None
+    secs = int(sleep_m.group(1))
+    check_intent = bool(re.search(
+        r"\b(ci|deploy|release|status|check|poll|pipeline|run)\b", cmd,
+        re.IGNORECASE))
+    sha = sha_hint or ""
+    m = re.search(r"\b([0-9a-f]{40}|[0-9a-f]{12})\b", cmd)
+    if m:
+        sha = m.group(1)
+    if not check_intent and not sha:
+        return None  # plain sleep, no continuation intent
+    cands = pending_operations(sha=sha) if sha else pending_operations()
+    if not cands and not sha:
+        return None
+    if len(cands) > 1:
+        return {"normalized": "AMBIGUOUS",
+                "candidates": [_rec_operation_id(c) for c in cands],
+                "raw_command_kind": "sleep_poll"}
+    rec = cands[0] if cands else None
+    if rec is not None:
+        norm = _canonical_from_record(rec, True, False, "sleep_poll_intent",
+                                      None, None)
+        norm["sha_for_wait"] = _rec_sha(rec)
+    elif sha:
+        op_type, evidence = derive_op_type(sha)
+        norm = {
+            "normalized": True, "action": "ContinuePendingOperation",
+            "operation_id": f"ci-{sha}", "op_type": op_type,
+            "authority": "ci_runs" if op_type in ("branch_ci",
+                                                  "pull_request_checks")
+                         else "deploy_qualify",
+            "deadline_secs": POLICY_DEADLINES[op_type],
+            "deadline_provenance": "GOVERNED_OPERATION_POLICY",
+            "model_deadline_ignored": True, "model_op_type_ignored": False,
+            "evidence": evidence, "sha_for_wait": sha,
+        }
+    else:
+        return None
+    norm["raw_command_kind"] = "sleep_poll"
+    norm["raw_sleep_seconds_suppressed"] = secs
+    return norm
+
+
+def emit_normalized_record(record: dict) -> None:
+    """Compact structured observability — one JSONL line per normalization.
+    No payloads, no secrets."""
+    try:
+        log = _SCRIPT_DIR.parent / "store" / "await_intent_log.jsonl"
+        log.parent.mkdir(parents=True, exist_ok=True)
+        slim = {k: record.get(k) for k in (
+            "normalized", "action", "operation_id", "op_type", "authority",
+            "deadline_secs", "deadline_provenance", "model_deadline_ignored",
+            "model_op_type_ignored", "matched_by", "evidence",
+            "raw_command_kind", "raw_sleep_seconds_suppressed", "reason",
+            "candidates") if k in record}
+        slim["at"] = __import__("time").strftime(
+            "%Y-%m-%dT%H:%M:%SZ", __import__("time").gmtime())
+        with open(log, "a") as f:
+            f.write(json.dumps(slim, sort_keys=True) + "\n")
+    except OSError:
+        pass
@@ -0,0 +1,74 @@
+#!/usr/bin/env python3
+"""await_intent_cli.py — CLI surface for the intent normalization layer.
+
+Used by the await-operation harness tool so the MODEL's scheduling hints
+never reach the wait engine. Prints one JSON line.
+"""
+from __future__ import annotations
+
+import argparse
+import json
+import sys
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).resolve().parent))
+import await_intent  # noqa: E402
+
+
+def main() -> int:
+    ap = argparse.ArgumentParser(description=__doc__)
+    sub = ap.add_subparsers(dest="cmd", required=True)
+    n = sub.add_parser("normalize", help="normalize a continuation intent")
+    n.add_argument("--sha", default="")
+    n.add_argument("--operation-id", default="")
+    n.add_argument("--task-id", default="",
+                   help="§9 task binding: scope pending resolution to "
+                        "operations registered for this task")
+    n.add_argument("--model-op-type", default=None)
+    n.add_argument("--model-deadline", default=None)
+    n.add_argument("--user-deadline", default=None,
+                   help="USER_EXPLICIT deadline (authoritative)")
+    n.add_argument("--contract-deadline", default=None,
+                   help="EXTERNAL_CONTRACT deadline (authoritative)")
+    rb = sub.add_parser("route-bash",
+                        help="route a Bash command expressing wait intent")
+    rb.add_argument("--command", required=True)
+    rb.add_argument("--sha-hint", default="")
+    args = ap.parse_args()
+
+    if args.cmd == "route-bash":
+        record = await_intent.normalize_bash(args.command,
+                                             sha_hint=args.sha_hint)
+        if record is None:
+            print(json.dumps({"normalized": False,
+                              "reason": "no external continuation intent"}))
+            return 0
+        if record.get("normalized") is True:
+            # the compiled wait needs the SHA: derive from operation_id
+            oid = str(record.get("operation_id", ""))
+            if oid.startswith("ci-"):
+                record["sha_for_wait"] = oid[3:]
+        await_intent.emit_normalized_record(record)
+        print(json.dumps(record, sort_keys=True))
+        return 0
+
+    action = {"kind": "await_operation", "sha": args.sha,
+              "operation_id": args.operation_id,
+              "op_type": args.model_op_type}
+    model_deadline = (int(args.model_deadline)
+                      if args.model_deadline else None)
+    user_deadline = (int(args.user_deadline)
+                     if args.user_deadline else None)
+    contract_deadline = (int(args.contract_deadline)
+                         if args.contract_deadline else None)
+    record = await_intent.normalize_action(
+        action, model_deadline=model_deadline,
+        user_deadline=user_deadline, contract_deadline=contract_deadline,
+        task_id=args.task_id)
+    await_intent.emit_normalized_record(record)
+    print(json.dumps(record, sort_keys=True))
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
@@ -0,0 +1,588 @@
+#!/usr/bin/env python3
+"""ci_event_wait.py — T16 §16C/§16D: event-driven CI qualification wait.
+
+Replaces the deploy.yml sleep-30 → query → repeat polling loop with the
+EXISTING W/R event bridge (src/rune/event_wait.py — the SAME primitive the
+builder adapter uses; §16E: one bridge, adapters differ only in identity
+and authority probe).
+
+Flow (mandate §16C):
+
+    push / CI workflow starts
+        ↓
+    register correlation  (provider="gitea", event_type="ci_completed",
+                           run_ref="ci-<sha>", sha=<sha>)
+        ↓
+    WAITING_FOR_EVENT     (await_resume_event: select() on FIFO, no floor)
+        ↓
+    ci.yml final job POSTs /process-event  (the WAKE)
+        ↓
+    daemon matches → resume_cmd fires → FIFO token → controller resumes
+        ↓
+    canonical Gitea state INDEPENDENTLY re-read (deploy_qualify.py verdict)
+        ↓
+    verdict
+
+The event is a WAKE, never truth. The final verdict always comes from
+deploy_qualify.py (required-checks context + exact-SHA ci.yml run, both
+legs). Watchdog deadline is a FAILURE DETECTOR, never the scheduler:
+
+  - event arrives → authority re-read; pending → bounded re-arm until the
+    overall deadline (ack-window semantics, same as the daemon's wake sweep).
+  - event lost/suppressed → watchdog expires → ONE-SHOT authority query:
+    terminal verdict → recovered truthfully (no invented wait);
+    CI still running (non-terminal Gitea runs for the SHA) → re-arm;
+    nothing alive and no verdict → truthful failure (deployment refused).
+
+Exit codes: 0 = qualified (success), 1 = refused (failure/timeout).
+Output: one JSON line with the §16D latency-proof timestamps.
+"""
+from __future__ import annotations
+
+import json
+import os
+import subprocess
+import sys
+import time
+from pathlib import Path
+
+_SCRIPT_DIR = Path(__file__).resolve().parent
+sys.path.insert(0, str(_SCRIPT_DIR))
+sys.path.insert(0, str(_SCRIPT_DIR.parent / "src"))
+
+sys.path.insert(0, str(_SCRIPT_DIR))
+from rune import event_wait as ew  # noqa: E402
+import completion_file as cf  # noqa: E402
+
+REPO_ROOT = _SCRIPT_DIR.parent
+QUALIFY_SCRIPT = _SCRIPT_DIR / "deploy_qualify.py"
+
+
+def authority_verdict(sha: str = "") -> str:
+    """ONE-SHOT canonical Gitea verdict via deploy_qualify.py — the only
+    completion truth. Never parses the event payload.
+
+    H-PATCH H4: the target sha is passed EXPLICITLY (as the subprocess's
+    --sha argument) — ambient DEPLOY_SHA can no longer change which
+    operation is verified. The env var remains only as a CLI backward
+    compat for deploy_qualify.py invoked standalone."""
+    cmd = [sys.executable, str(QUALIFY_SCRIPT)]
+    if sha:
+        cmd += ["--sha", sha]
+    proc = subprocess.run(
+        cmd,
+        capture_output=True, text=True, timeout=120,
+        env={k: v for k, v in os.environ.items() if k != "DEPLOY_SHA"}
+        if sha else dict(os.environ, PYTHONPATH=""),
+    )
+    out = (proc.stdout or "").strip().splitlines()
+    return (out[-1] if out else "pending").strip()
+
+
+# ── T21.5 §4: operation-specific completion authority ──────────────────────
+#
+# T21.4 root cause: the wait loop had ONE hard-coded authority
+# (deploy qualification = master-ref CI) while operations waited on
+# branch CI, PR checks, or merges. Wrong authority => nonterminal
+# returns => the MODEL had to re-arm (the generator of the sleeps).
+# The wait verifier MUST match the operation being waited for.
+
+OPERATION_SPECS = {
+    # op_type: {
+    #   workflow_path: Actions run path filter (None = any),
+    #   head_branches: head-branch filter (None = any),
+    #   authority: which canonical verifier decides completion,
+    #   event_types: expected wake event types,
+    #   terminal: terminal verdicts, failure: failure verdicts,
+    # }
+    "branch_ci": {
+        "workflow_path": "ci.yml",
+        "head_branches": None,
+        "authority": "ci_runs",          # any completed run for this SHA
+        "event_types": ("ci_completed",),
+        "terminal": ("success",),
+        "failure": ("failure", "cancelled"),
+    },
+    "pull_request_checks": {
+        "workflow_path": None,           # all PR-triggered runs for SHA
+        "head_branches": None,
+        "authority": "ci_runs",
+        "event_types": ("ci_completed",),
+        "terminal": ("success",),
+        "failure": ("failure", "cancelled"),
+    },
+    "master_ci": {
+        "workflow_path": "ci.yml",
+        "head_branches": ("master",),
+        "authority": "deploy_qualify",   # full two-leg qualification
+        "event_types": ("ci_completed",),
+        "terminal": ("success",),
+        "failure": ("failure", "cancelled"),
+    },
+    "deploy": {
+        "workflow_path": "deploy.yml",
+        "head_branches": ("master",),
+        "authority": "deploy_qualify",
+        "event_types": ("ci_completed", "deploy_completed"),
+        "terminal": ("success",),
+        "failure": ("failure", "cancelled"),
+    },
+    "release": {
+        "workflow_path": None,
+        "head_branches": ("master",),
+        "authority": "deploy_qualify",
+        "event_types": ("ci_completed", "release_completed"),
+        "terminal": ("success",),
+        "failure": ("failure", "cancelled"),
+    },
+}
+
+DEFAULT_OP_TYPE = "master_ci"
+
+
+def normalize_verdict(verdict: str) -> str:
+    """T21.5 §18 audit fix: pending-class verdicts are ALL nonterminal.
+    deploy_qualify emits "pending (probe failed)" for a transient probe
+    failure — that is pending, never a terminal answer. The wait loop
+    must not return control to the model over a string variant."""
+    v = (verdict or "").strip().lower()
+    if v.startswith("pending"):
+        return "pending"
+    return (verdict or "").strip()
+
+
+def op_spec(op_type: str) -> dict:
+    spec = OPERATION_SPECS.get(op_type)
+    if spec is None:
+        raise ValueError(
+            f"unknown op_type {op_type!r} — must be one of "
+            f"{sorted(OPERATION_SPECS)} (the wait verifier must match "
+            "the operation being waited for)")
+    return spec
+
+
+def op_authority_verdict(op_type: str, sha: str) -> str:
+    """Canonical completion verdict FOR THIS OPERATION TYPE.
+
+    ci_runs        — the op's own Actions runs for this exact SHA: every
+                     required workflow run completed; terminal iff every
+                     run of the op's workflow set concluded success
+                     (branch/PR ops are their own truth; deploy
+                     qualification is NOT their authority).
+    deploy_qualify — the two-leg deploy qualification (master CI +
+                     required-checks) via deploy_qualify.py.
+    """
+    spec = op_spec(op_type)
+    authority = spec["authority"]
+    if authority == "deploy_qualify":
+        v = normalize_verdict(authority_verdict(sha))
+        if v != "pending":
+            return v
+        # E-PATCH resilience: Gitea's combined commit-status endpoint can
+        # 404 for the newest commit (observed live on auto-version bump
+        # commits — the status object materializes late/never). The runs
+        # ARE the same evidence the two legs read; fall back to the
+        # operation-scoped runs verdict so authority stays resolvable.
+        if _status_endpoint_unreadable(sha):
+            return _ci_runs_verdict(spec, sha)
+        return v
+    if authority == "ci_runs":
+        return _ci_runs_verdict(spec, sha)
+    raise ValueError(f"unknown authority {authority!r} for {op_type!r}")
+
+
+def _status_endpoint_unreadable(sha: str) -> bool:
+    """True when the combined commit-status endpoint 404s for the SHA
+    (index lag / never-materialized status object) — the documented
+    fallback trigger, verified against the API once per call."""
+    try:
+        sys.path.insert(0, str(_SCRIPT_DIR))
+        import gitea_auth  # noqa: E402
+        st, _body = gitea_auth.api(f"/commits/{sha}/status")
+        return st == 404
+    except Exception:  # noqa: BLE001
+        return False
+
+
+def _ci_runs_verdict(spec: dict, sha: str) -> str:
+    """Operation-scoped Actions-run verdict (import the API client from
+    deploy_qualify to keep ONE Gitea client)."""
+    sys.path.insert(0, str(_SCRIPT_DIR))
+    import deploy_qualify as dq  # noqa: E402
+    import gitea_auth  # noqa: E402
+    server = os.environ.get("GITEA_SERVER",
+                            getattr(gitea_auth, "DEFAULT_SERVER", "")).rstrip("/")
+    repo = os.environ.get("GITEA_REPO",
+                          getattr(gitea_auth, "DEFAULT_REPO", ""))
+    sha_env = sha  # H-PATCH: explicit parameter is the only authority
+    token = gitea_auth.get_token() if gitea_auth.has_token() else ""
+    if not (server and repo and sha_env and token):
+        return "pending"  # cannot verify -> not verified (fail closed)
+    runs = dq._fetch_runs(server, repo, sha_env, token)
+    if runs is None:
+        return "pending"  # probe failure -> keep waiting
+    wfp = spec["workflow_path"]
+    if wfp is not None:
+        runs = [r for r in runs
+                if str(r.get("path", "")).startswith(wfp + "@")]
+    branches = spec["head_branches"]
+    if branches is not None:
+        runs = [r for r in runs
+                if str(r.get("head_branch", "")) in branches]
+    if not runs:
+        return "pending"  # no runs visible yet -> not terminal
+    for r in runs:
+        if r.get("status") != "completed":
+            return "pending"  # still running
+        if r.get("conclusion") not in ("success", "skipped"):
+            return "failure"  # any non-success conclusion
+    return "success"
+
+
+def op_ci_alive(op_type: str, sha: str) -> bool:
+    """Liveness probe FOR THIS OPERATION TYPE: any non-terminal run in
+    the op's run set (failure DETECTOR, never the scheduler)."""
+    sys.path.insert(0, str(_SCRIPT_DIR))
+    import deploy_qualify as dq  # noqa: E402
+    import gitea_auth  # noqa: E402
+    spec = op_spec(op_type)
+    server = os.environ.get("GITEA_SERVER",
+                            getattr(gitea_auth, "DEFAULT_SERVER", "")).rstrip("/")
+    repo = os.environ.get("GITEA_REPO",
+                          getattr(gitea_auth, "DEFAULT_REPO", ""))
+    sha_env = sha  # H-PATCH: explicit parameter is the only authority
+    token = gitea_auth.get_token() if gitea_auth.has_token() else ""
+    if not (server and repo and sha_env and token):
+        return False
+    runs = dq._fetch_runs(server, repo, sha_env, token)
+    if not runs:
+        return False
+    wfp = spec["workflow_path"]
+    if wfp is not None:
+        runs = [r for r in runs
+                if str(r.get("path", "")).startswith(wfp + "@")]
+    branches = spec["head_branches"]
+    if branches is not None:
+        runs = [r for r in runs
+                if str(r.get("head_branch", "")) in branches]
+    terminal = {"success", "failure", "cancelled", "skipped"}
+    return any(str(r.get("status", "")).lower() not in terminal
+               for r in runs)
+
+
+def authority_after_wake(op_type: str = DEFAULT_OP_TYPE, sha: str = "",
+                         max_tries: int = 3, gap_secs: int = 4) -> str:
+    """(sha is EXPLICIT — ambient env never selects the target.)"""
+    """Post-wake authority read. The daemon's own wake sweep uses a
+    bounded event-triggered retry (4s ×3) for the job-finished-microseconds-
+    ago ack window; a verdict read here can hit the same window. This is a
+    bounded schedule, not a polling loop — 3 reads max, then back to the
+    event wait / watchdog."""
+    verdict = "pending"
+    for attempt in range(max_tries):
+        verdict = op_authority_verdict(op_type, sha)
+        if verdict != "pending":
+            return verdict
+        if attempt < max_tries - 1:
+            time.sleep(gap_secs)
+    return verdict
+
+
+def ci_alive() -> bool:
+    """LEGACY liveness probe (master-ref deploy runs) — kept for callers
+    predating op specs; run() uses op_ci_alive(op_type, sha)."""
+    return op_ci_alive(DEFAULT_OP_TYPE, os.environ.get("DEPLOY_SHA", ""))
+
+
+def _set_cognition_state(op: str, state: str) -> None:
+    """E-PATCH §12: stamp cognitive quiescence on the durable record so a
+    fresh process knows cognition is parked (WAITING_EXTERNAL) vs resumable
+    (RESUMABLE). Best-effort: observability never breaks the wait."""
+    try:
+        rec = ew.operation_wait(op)
+        if rec is None:
+            return
+        wait_file = ew.wr_root() / "waits" / f"{rec.get('wait_id')}.json"
+        if not wait_file.is_file():
+            return
+        data = json.loads(wait_file.read_text())
+        data.setdefault("ext", {})["continuation_state"] = state
+        wait_file.write_text(json.dumps(data, indent=1, sort_keys=True))
+    except (OSError, ValueError):
+        pass
+
+
+def run(*, sha: str, deadline_secs: int,
+        watchdog_secs: int = 300,
+        op_type: str = DEFAULT_OP_TYPE,
+        task_id: str = "",
+        daemon_url: str = ew.DEFAULT_DAEMON_URL) -> int:
+    """T21.5 §3: ONE LOGICAL AWAIT.
+
+    One logical external wait is one invocation of this function. All
+    nonterminal conditions (event-seen-authority-pending, watchdog expiry
+    with live runs, bounded authority-retry exhausted) re-arm INTERNALLY —
+    control returns to the caller ONLY on a terminal verdict or on real
+    deadline expiry. Watchdogs are failure detectors inside this
+    abstraction, never the scheduler and never completion truth.
+    """
+    spec = op_spec(op_type)          # unknown op_type refused upfront
+    op = f"ci-{sha}"
+    # A-PATCH: the authority subprocess reads DEPLOY_SHA from the env;
+    # python-level run() callers (tool normalization flow) must export it.
+    if sha and not os.environ.get("DEPLOY_SHA"):
+        os.environ["DEPLOY_SHA"] = sha
+    # H-PATCH: resolve short SHAs to full 40-char — _fetch_runs and
+    # deploy_qualify match on the FULL sha; a 12-char prefix would
+    # silently fail equality and park the waiter until deadline
+    # (the 962e351 root cause).
+    if sha and len(sha) < 40:
+        try:
+            full = subprocess.run(
+                ["git", "-C", str(_SCRIPT_DIR.parent), "rev-parse", sha],
+                capture_output=True, text=True, timeout=10)
+            if full.returncode == 0:
+                sha = full.stdout.strip()
+                os.environ["DEPLOY_SHA"] = sha
+        except (OSError, subprocess.TimeoutExpired):
+            pass
+    t0 = time.time()
+    # REAL_ASYNC_BRIDGE: FIFO is replaced by completion-file inotify.
+    # The completion file write is the wake signal; inotify is the
+    # kernel-level zero-latency, zero-CPU wake primitive.
+    # Event replay (T16 §16C live finding): on a serialized runner the
+    # completion event usually arrives BEFORE the waiter registers.
+    # A matching durable event means the wake already happened — skip
+    # the rendezvous entirely and go straight to the authority read.
+    prior = ew.find_matching_event("gitea", "ci_completed", op)
+    if prior is not None and prior.get("exit_code") is not None:
+        rec = ew.register_process_wait(
+            op, project="rune",
+            objective=f"event-driven {op_type} qualification for {sha}",
+            watched_for="Gitea CI completion event (replayed from the "
+                        "durable event log)",
+            provider="gitea", event_type="ci_completed", sha=sha,
+            watchdog_secs=min(watchdog_secs, deadline_secs),
+            continuation_state="RESUMABLE",  # event already arrived
+            op_type=op_type,
+            authority=spec["authority"])
+        registered_at = rec["created_at"]
+        event_at = time.strftime(
+            "%Y-%m-%dT%H:%M:%SZ", time.gmtime())
+        resumed_at = event_at
+        # the event is a WAKE, never truth — canonical authority decides
+        verdict = normalize_verdict(authority_after_wake(op_type, sha))
+        if verdict == "pending" and not op_ci_alive(op_type, sha):
+            verdict = normalize_verdict(op_authority_verdict(op_type, sha))
+        mode = ("event_replayed" if verdict != "pending"
+                else "authority_unresolved")
+        verdict_at = time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime())
+        out = {
+            "sha": sha,
+            "op_type": op_type,
+            "authority": spec["authority"],
+            "registered_at": registered_at,
+            "event_received_at": event_at,
+            "controller_resumed_at": resumed_at,
+            "verdict_at": verdict_at,
+            "mode": mode,
+            "rearm_reason": "completion event replayed from durable log "
+                            "(arrived before registration)",
+            "verdict": verdict,
+            "wall_secs": round(time.time() - t0, 3),
+        }
+        print(json.dumps(out))
+        return 0 if verdict == "success" else 1
+    rec = ew.register_process_wait(
+        op, project="rune",
+        objective=f"event-driven {op_type} qualification for {sha}",
+        watched_for="Gitea CI completion event (wake; verdict re-read "
+                    f"from canonical {spec['authority']} authority)",
+        provider="gitea", event_type="ci_completed", sha=sha,
+        watchdog_secs=min(watchdog_secs, deadline_secs),
+        continuation_state="WAITING_EXTERNAL",
+        op_type=op_type,
+        authority=spec["authority"],
+        task_id=task_id,  # §9: bind the operation to the calling task
+        resume_mode="self")  # §3: waiter owns its continuation via
+                             # completion-file/inotify; the daemon has NO
+                             # session-delivery role for this wait
+    registered_at = rec["created_at"]
+    _set_cognition_state(op, "WAITING_EXTERNAL")  # cognition parked
+    deadline = time.monotonic() + deadline_secs
+    verdict = "pending"
+    mode = "event"
+    event_at = ""
+    resumed_at = ""
+    rearm_reason = ""
+    rearm_count = 0
+    while time.monotonic() < deadline:
+            remaining = max(1, int(deadline - time.monotonic()))
+
+            # REAL_ASYNC_BRIDGE: block on inotify for the completion file.
+            # Zero CPU, zero arbitrary timer. The inotify timeout is a
+            # FAILURE DETECTOR — if no event arrives within the watchdog
+            # window, we check the authority directly (lost-event recovery).
+            # The deadline is containment only.
+            watch_timeout = min(remaining, watchdog_secs)
+            result = cf.wait_for_completion(
+                op, timeout_secs=watch_timeout)
+
+            # REAL_ASYNC_BRIDGE: check durable event log AND wait record
+            # status on EVERY cycle — the daemon may have matched and
+            # transitioned the record via /process-event.
+            wr = ew.operation_wait(op)
+            if wr and wr.get("status") in ("COMPLETED", "FAILED",
+                                           "READY_TO_RESUME"):
+                event_at = event_at or time.strftime(
+                    "%Y-%m-%dT%H:%M:%SZ", time.gmtime())
+                resumed_at = time.strftime("%Y-%m-%dT%H:%M:%SZ",
+                                           time.gmtime())
+                verdict = normalize_verdict(
+                    authority_after_wake(op_type, sha))
+                if verdict != "pending":
+                    mode = "event"
+                    break
+            prior = ew.find_matching_event("gitea", "ci_completed", op)
+            if prior is not None and prior.get("exit_code") is not None:
+                event_at = event_at or time.strftime(
+                    "%Y-%m-%dT%H:%M:%SZ", time.gmtime())
+                resumed_at = time.strftime("%Y-%m-%dT%H:%M:%SZ",
+                                           time.gmtime())
+                verdict = normalize_verdict(
+                    authority_after_wake(op_type, sha))
+                if verdict != "pending":
+                    mode = "event"
+                    break
+            if result is not None:
+                # completion file exists: event arrived
+                event_at = event_at or time.strftime(
+                    "%Y-%m-%dT%H:%M:%SZ", time.gmtime())
+                resumed_at = time.strftime("%Y-%m-%dT%H:%M:%SZ",
+                                           time.gmtime())
+                verdict = normalize_verdict(
+                    authority_after_wake(op_type, sha))
+                if verdict != "pending":
+                    mode = "event"
+                    break
+                rearm_count += 1
+                rearm_reason = ("completion file seen; authority "
+                                "pending — re-subscribing")
+                continue
+
+            # inotify timeout: no event within watchdog window
+            # FAILURE DETECTOR: check authority directly
+            verdict = normalize_verdict(
+                op_authority_verdict(op_type, sha))
+            if verdict != "pending":
+                mode = "watchdog_authority_recovered"
+                break
+            rearm_count += 1
+            rearm_reason = ("watchdog expiry; op pending (not started "
+                            "or lagging) — re-subscribing inotify")
+            continue
+
+    # REAL_ASYNC_BRIDGE: no FIFO to close. The completion file is durable
+    # and survives restarts. Lease release is not needed (no lease system
+    # for file-based delivery).
+
+
+    if verdict == "pending":
+        if mode == "event":
+            mode = "deadline_pending"
+        verdict = normalize_verdict(op_authority_verdict(op_type, sha))  # final one-shot truth
+    _set_cognition_state(op, "RESUMABLE" if verdict != "pending"
+                         else "RECOVERY_REQUIRED")
+    verdict_at = time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime())
+    # §2 STATE OWNERSHIP: the WAITER owns the OPERATION domain. On a
+    # terminal authority verdict, stamp the durable record terminal with
+    # the evidence — the daemon's session-delivery machinery must never
+    # (and can no longer) overwrite this truth. Also updates the parked
+    # continuation_state so fresh-process recovery sees reality.
+    try:
+        rec2 = ew.operation_wait(op) or rec
+        if verdict != "pending":
+            rec2["status"] = "COMPLETED" if verdict == "success" else "FAILED"
+            rec2["verdict"] = verdict
+            rec2["resume_evidence"] = {
+                "kind": "authority_verified_terminal",
+                "verdict": verdict,
+                "mode": mode,
+                "authority": spec["authority"],
+                "sha": sha,
+                "op_type": op_type,
+                "verdict_at": verdict_at,
+            }
+        rec2["ext"] = {**rec2.get("ext", {}),
+                       "continuation_state": ("RESUMABLE" if verdict != "pending"
+                                              else "RECOVERY_REQUIRED")}
+        ew._load_waitstore().save(rec2)
+    except Exception as _e:  # noqa: BLE001 — record stamping must never corrupt the envelope (the envelope + authority remain the truth)
+        print(f"ci_event_wait: terminal record stamp failed: {_e}",
+              file=sys.stderr)
+    out = {
+        "sha": sha,
+        "op_type": op_type,
+        "authority": spec["authority"],
+        "registered_at": registered_at,
+        "event_received_at": event_at,
+        "controller_resumed_at": resumed_at,
+        "verdict_at": verdict_at,
+        "mode": mode,
+        "rearm_reason": rearm_reason,
+        "internal_rearm_count": rearm_count,
+        "verdict": verdict,
+        "wall_secs": round(time.time() - t0, 3),
+    }
+    print(json.dumps(out))
+    return 0 if verdict == "success" else 1
+
+KNOWN_FLAGS = ("--sha", "--deadline", "--op-type", "--watchdog", "--task-id")
+
+
+def main() -> int:
+    """§5 CLI CONTRACT: no silent argument loss. Every flag is either
+    explicitly supported (and takes effect) or loudly rejected — an
+    acceptance envelope must reflect the configuration actually used."""
+    sha = os.environ.get("DEPLOY_SHA", "")
+    args = sys.argv[1:]
+    deadline = 900
+    op_type = DEFAULT_OP_TYPE
+    watchdog = 300
+    task_id = ""
+    i = 0
+    while i < len(args):
+        a = args[i]
+        if a == "--sha":
+            sha = args[i + 1]; i += 2
+        elif a == "--deadline":
+            deadline = int(args[i + 1]); i += 2
+        elif a == "--op-type":
+            op_type = args[i + 1]; i += 2
+        elif a == "--watchdog":
+            watchdog = int(args[i + 1]); i += 2
+        elif a == "--task-id":
+            task_id = args[i + 1]; i += 2
+        elif a.startswith("--"):
+            print(json.dumps({"error": f"unknown flag {a!r}",
+                              "known_flags": list(KNOWN_FLAGS),
+                              "contract": "no silent argument loss"}))
+            return 2
+        else:
+            print(json.dumps({"error": f"unexpected positional {a!r}",
+                              "known_flags": list(KNOWN_FLAGS)}))
+            return 2
+    if not sha:
+        print(json.dumps({"error": "no sha (DEPLOY_SHA or --sha required)"}))
+        return 1
+    # A-PATCH fix: run()'s authority subprocess (deploy_qualify.py) reads
+    # DEPLOY_SHA from the environment — a CLI `--sha` consumer previously
+    # got a silent "pending" verdict forever because the var was never
+    # exported. Export it so the canonical authority can resolve.
+    os.environ["DEPLOY_SHA"] = sha
+    return run(sha=sha, deadline_secs=deadline, op_type=op_type,
+               watchdog_secs=watchdog, task_id=task_id)
+
+
+if __name__ == "__main__":
+    sys.exit(main())
@@ -0,0 +1,195 @@
+"""completion_file.py — REAL_ASYNC_BRIDGE: durable completion-file delivery
+with inotify wake. Replaces FIFO+lease for process-wait delivery.
+
+DESIGN:
+  Producer (bridge/daemon/notify):
+    1. Validates operation identity
+    2. Atomically writes a completion file:
+         resumes/<operation_id>.done
+       containing the event/verdict JSON (write to tmp, rename for atomicity)
+
+  Waiter (ci_event_wait / any continuation consumer):
+    1. Checks if the completion file already exists (event-before-waiter)
+    2. If not: registers inotify watch on the resumes/ directory
+    3. Re-checks after watch registration (closes the registration race)
+    4. Blocks in kernel on inotify — zero CPU, zero timer
+    5. On file-created event: reads the file, verifies authority, returns
+
+  The event is a WAKE SIGNAL. Authority still decides terminal truth.
+  The completion file is durable — survives daemon and waiter restarts.
+
+WHY NOT FIFO:
+  FIFO is one-shot (token consumed), requires lease management, and
+  fails under session-busy guards. A completion file is atomic, durable,
+  and works across restarts. inotify wakes the process with zero CPU
+  cost and zero latency between file creation and wake.
+
+WHY NOT POLLING:
+  Polling introduces an arbitrary interval. inotify is a kernel callback
+  that fires exactly when the filesystem changes — the latency tracks
+  the real completion event, not a configured interval.
+"""
+from __future__ import annotations
+
+import json
+import os
+import select
+import struct
+import time
+from pathlib import Path
+
+def _wr_bridge() -> Path:
+    # WR_ROOT = daemon/infra convention; RUNE_WR_ROOT = repo/scripts
+    # convention. Both honored (WR_ROOT wins) so the daemon, the waiter,
+    # and any isolated test root always resolve the SAME directory.
+    root = (os.environ.get("WR_ROOT")
+            or os.environ.get("RUNE_WR_ROOT")
+            or "/mnt/workdrive/wr-bridge")
+    return Path(root)
+
+def _resumes_dir() -> Path:
+    return _wr_bridge() / "resumes"
+
+# inotify constants (Linux)
+IN_CREATE = 0x100
+IN_CLOSE_WRITE = 0x8
+IN_MOVED_TO = 0x80
+
+
+def completion_path(operation_id: str) -> Path:
+    return _resumes_dir() / f"{operation_id}.done"
+
+
+def write_completion(operation_id: str, event: dict) -> Path:
+    """Atomically write a durable completion file. The rename(2) ensures
+    the waiter never sees a partial file."""
+    _resumes_dir().mkdir(parents=True, exist_ok=True)
+    target = completion_path(operation_id)
+    tmp = target.parent / f".{target.name}.tmp.{os.getpid()}"
+    tmp.write_text(json.dumps(event, sort_keys=True) + "\n")
+    os.rename(str(tmp), str(target))  # atomic on same filesystem
+    os.chmod(str(target), 0o666)  # cross-user bridge
+    return target
+
+
+def read_completion(operation_id: str) -> dict | None:
+    """Read a durable completion file. Returns None if absent."""
+    p = completion_path(operation_id)
+    if not p.is_file():
+        return None
+    try:
+        return json.loads(p.read_text())
+    except (OSError, ValueError):
+        return None
+
+
+def completion_exists(operation_id: str) -> bool:
+    return completion_path(operation_id).is_file()
+
+
+def _inotify_init_and_add_watch(directory: str) -> int:
+    """Create an inotify instance and add a watch on the directory.
+    Returns the inotify fd."""
+    libc = __import__("ctypes").CDLL(None, use_errno=True)
+    fd = libc.inotify_init()
+    if fd < 0:
+        raise OSError("inotify_init failed")
+    # IN_CREATE | IN_MOVED_TO: catch both direct creation and atomic rename
+    mask = IN_CREATE | IN_MOVED_TO
+    wd = libc.inotify_add_watch(fd, directory.encode(), mask)
+    if wd < 0:
+        os.close(fd)
+        raise OSError(f"inotify_add_watch failed for {directory}")
+    return fd
+
+
+def _inotify_read(fd: int, timeout_secs: float | None = None) -> list[str]:
+    """Read pending inotify events. Returns list of filenames that were
+    created/moved-to. Blocks until at least one event or timeout."""
+    poll = select.poll()
+    poll.register(fd, select.POLLIN)
+    events = []
+    buf = os.read(fd, 4096)
+    pos = 0
+    while pos < len(buf):
+        # struct: wd(i32), mask(u32), cookie(u32), name_len(u32)
+        wd, mask, cookie, name_len = struct.unpack_from("iIII", buf, pos)
+        pos += 16
+        name = buf[pos:pos + name_len].rstrip(b"\x00").decode("utf-8", "replace")
+        pos += name_len
+        if mask & (IN_CREATE | IN_MOVED_TO):
+            events.append(name)
+    return events
+
+
+def wait_for_completion(operation_id: str, *,
+                        timeout_secs: float | None = None,
+                        poll_callback=None) -> dict | None:
+    """Block until a completion file exists for this operation.
+
+    Uses inotify for zero-latency, zero-CPU wake. Falls back to
+    polling_callback (if provided) when inotify is unavailable.
+
+    The completion file is a WAKE SIGNAL. The caller must still verify
+    canonical authority after reading it.
+
+    Returns the event dict from the completion file, or None on timeout.
+    """
+    target = completion_path(operation_id)
+
+    # 1. check-before-watch (event may have arrived before we started)
+    existing = read_completion(operation_id)
+    if existing is not None:
+        return existing
+
+    _resumes_dir().mkdir(parents=True, exist_ok=True)
+
+    try:
+        fd = _inotify_init_and_add_watch(str(_resumes_dir()))
+    except OSError:
+        # inotify unavailable: fall back to polling callback
+        if poll_callback:
+            return poll_callback(operation_id, timeout_secs)
+        return None
+
+    try:
+        # 2. re-check after watch registration (closes the race between
+        #    the initial check and the watch being active)
+        existing = read_completion(operation_id)
+        if existing is not None:
+            return existing
+
+        # 3. block in kernel on inotify — zero CPU, exact latency
+        import select as _select
+        poll = _select.poll()
+        poll.register(fd, _select.POLLIN)
+        deadline = (time.monotonic() + timeout_secs) if timeout_secs else None
+
+        while True:
+            remaining = None
+            if deadline is not None:
+                remaining = max(0.1, deadline - time.monotonic())
+                if remaining <= 0.1:
+                    return None  # timeout
+            events = poll.poll(remaining * 1000 if remaining else None)
+            if not events:
+                continue  # poll timeout, re-check
+            names = _inotify_read(fd)
+            if target.name in names:
+                # 4. file appeared — read it
+                existing = read_completion(operation_id)
+                if existing is not None:
+                    return existing
+                # partial/corrupt: keep watching
+    finally:
+        os.close(fd)
+    return None
+
+
+def clear_completion(operation_id: str) -> None:
+    """Remove a consumed completion file (idempotent cleanup)."""
+    p = completion_path(operation_id)
+    try:
+        p.unlink()
+    except OSError:
+        pass
@@ -0,0 +1,123 @@
+"""construction_guards.py — deterministic guards for observed process
+defects (SCAFFOLD-TRANCHE-006 §21-22).
+
+Two real defect classes from recent sessions:
+
+A. DURABLE-STORE MUTATION WITHOUT PERSISTENCE (Tranche 5: closure script
+   mutated the contract in memory, never called save(); the next read saw
+   stale state). Guard: a durable-boundary operation verifies by
+   RELOAD/READBACK, not by successful in-memory mutation.
+
+B. BLIND SOURCE-EDIT NO-OP (multiple sessions: python str.replace() on
+   ruff-reformatted files silently matched nothing; the script reported
+   success while changing nothing). Guard: an automated edit verifies its
+   POSTCONDITION — expected text present / old text absent — and a
+   zero-match replacement is an explicit, truthful no-op, never success.
+
+Not a generic editor framework. Two small, deterministic postcondition
+checks at the boundaries where the defects actually occurred.
+"""
+
+from __future__ import annotations
+
+import json
+from pathlib import Path
+
+__all__ = ["verify_persisted", "assert_persisted", "apply_source_edit", "EditNoOp", "EditPostconditionFailed"]
+
+
+class EditNoOp(Exception):
+    """The edit matched zero targets — an explicit, truthful no-op."""
+
+
+class EditPostconditionFailed(Exception):
+    """The edit claimed success but the expected postcondition is absent."""
+
+
+# ── Guard A: durable mutation verified by readback ──────────────────────────
+
+
+def verify_persisted(path: Path | str, selector, expect) -> bool:
+    """Reload the durable store from disk and check the expectation against
+    the RELOADED state. `selector(reloaded_doc) -> value`; pass == value
+    equals `expect` (or expect(value) is truthy when callable).
+
+    The reloaded document is parsed as JSON when the file is JSON-shaped;
+    otherwise the raw text is passed to the selector (plain-text stores).
+
+    This is the Tranche 5 defect guard: an in-memory mutation that was
+    never save()d fails here, because the disk state is what gets read.
+    """
+    raw = Path(path).read_text(encoding="utf-8")
+    try:
+        doc = json.loads(raw)
+    except ValueError:
+        doc = raw
+    value = selector(doc)
+    if callable(expect):
+        return bool(expect(value))
+    return value == expect
+
+
+def assert_persisted(path: Path | str, selector, expect, *, what: str = "durable state") -> None:
+    """verify_persisted that raises on mismatch — for use right after a
+    save() at durable boundaries."""
+    if not verify_persisted(path, selector, expect):
+        raise AssertionError(
+            f"persistence verification FAILED for {what}: reloaded state "
+            f"does not match expectation — the mutation may not have been "
+            f"saved (Tranche 5 defect class A)"
+        )
+
+
+# ── Guard B: source edit with postcondition ─────────────────────────────────
+
+
+def apply_source_edit(
+    text: str,
+    old: str,
+    new: str,
+    *,
+    expected_count: int | None = None,
+    postcondition: str | None = None,
+    allow_noop: bool = False,
+) -> str:
+    """Apply a source edit WITH postcondition verification (Guard B).
+
+    - zero matches: raises EditNoOp (truthful no-op) unless allow_noop=True
+    - expected_count given: match count must equal it
+    - postcondition given: the RESULT must contain it, else
+      EditPostconditionFailed
+    Returns the new text. Never claims success for a change that did not
+    happen."""
+    count = text.count(old)
+    if count == 0:
+        if allow_noop:
+            return text  # truthful, explicit no-op — caller records it
+        raise EditNoOp(
+            f"edit matched 0 targets — nothing changed (Tranche 5 defect "
+            f"class B: str.replace silently no-ops on reformatted files)"
+        )
+    if expected_count is not None and count != expected_count:
+        raise EditPostconditionFailed(f"edit matched {count} targets, expected {expected_count}")
+    result = text.replace(old, new)
+    if postcondition and postcondition not in result:
+        raise EditPostconditionFailed(
+            "postcondition text absent from result — edit did not achieve its expected effect"
+        )
+    return result
+
+
+def apply_source_edit_file(
+    path: Path | str, old: str, new: str, *, expected_count: int | None = None, postcondition: str | None = None
+) -> int:
+    """File-level convenience: read, apply_source_edit, write back, then
+    VERIFY BY READBACK (both guards composed). Returns match count."""
+    p = Path(path)
+    text = p.read_text(encoding="utf-8")
+    result = apply_source_edit(text, old, new, expected_count=expected_count, postcondition=postcondition)
+    p.write_text(result, encoding="utf-8")
+    # Guard A composed: the written file must contain the postcondition
+    if postcondition:
+        assert_persisted(p, lambda t: t, lambda t: postcondition in t, what=f"edit of {p.name}")
+    return text.count(old)
@@ -24,6 +24,11 @@ import sys
 import urllib.error
 import urllib.request
 from typing import Any
+import sys as _sys
+_SCRIPT_DIR = str(__import__('pathlib').Path(__file__).resolve().parent)
+if _SCRIPT_DIR not in _sys.path:
+    _sys.path.insert(0, _SCRIPT_DIR)
+import gitea_auth

 # Env-contract (#192 spec-gate finding 1): derived in-script, never required
 # from the caller. Guarded by tests/test_workflow_script_contract.py.
@@ -87,7 +92,7 @@ def main(argv: list[str] | None = None) -> int:
        print(json.dumps(payload))
        return 0

-    token = os.environ["GITEA_TOKEN"]
+    token = gitea_auth.get_token()

    # Derived URLs (explicit env still overrides — for direct/manual use).
    server_url = os.environ["RELEASE_SERVER_URL"]
@@ -0,0 +1,224 @@
+#!/usr/bin/env python3
+"""deploy_promotion.py — monotonic promotion authority for Rune deploys.
+
+SCAFFOLD contract DEPLOY-MONOTONIC-001. Root cause of the 2026-09-13
+deploy-order race (deploys 7355/7359): deploy_stage.py promoted
+unconditionally in job-finish order — no comparison against the currently
+deployed commit, no lock, no truthful outcome for a stale candidate.
+
+This module owns the PROMOTION decision. The invariant:
+
+    current never regresses along the canonical release branch.
+
+Ordering authority: GIT ANCESTRY on the canonical release branch
+(refs/heads/master). Rune's release flow is linear history — every deploy
+candidate is a master push (tranche commit or its auto-bump descendant).
+    candidate == deployed commit      -> NOOP_ALREADY_CURRENT (idempotent)
+    candidate is DESCENDANT of it     -> allowed (normal forward deploy)
+    candidate is ANCESTOR of it       -> STALE_SUPERSEDED (refuse; truthful)
+    no ancestry relationship          -> REFUSED_DIVERGENT (fail closed)
+    deployed commit unknown/unreadable-> REFUSED_UNKNOWN_CURRENT (fail closed)
+
+Mechanism (smallest that provides the invariant, per brief §4-5):
+    1. promotion-time stale check — ancestry comparison against the commit
+       current ACTUALLY points to at promotion time (not at job start);
+    2. an flock'd critical section around the read-compare-write of
+       current/last-good so two concurrent promotions cannot interleave
+       between the ancestry check and the symlink swap.
+Workflow-level concurrency is NOT relied on (brief §5): the fence lives at
+the promotion operation itself.
+
+Truthful outcomes (brief §8) — build/stage success != promotion success:
+    PROMOTED                candidate became current (verified below)
+    NOOP_ALREADY_CURRENT    same commit already deployed; nothing mutated
+    STALE_SUPERSEDED        a newer commit is already current; nothing mutated
+    REFUSED_DIVERGENT       ordering unprovable; fail closed
+    REFUSED_UNKNOWN_CURRENT deployed commit unreadable; fail closed
+    FAILED_*                verification/staging failures (in deploy_stage)
+
+A stale or refused outcome is a NON-ERROR terminal state for the release
+pipeline (the newer build simply won) but MUST NOT print "PROMOTED" and
+MUST NOT mutate current/last-good. deploy_stage's exit code and the
+workflow's post-deploy runtime verification enforce this: the workflow
+asserts current == candidate after a PROMOTED outcome, so a stale outcome
+makes the job report honestly rather than masquerade.
+"""
+
+from __future__ import annotations
+
+import fcntl
+import json
+import os
+import subprocess
+from pathlib import Path
+
+DEPLOY_ROOT = Path("/mnt/workdrive/rune-code")
+RELEASE_BRANCH = "refs/heads/master"
+
+OUTCOME_PROMOTED = "PROMOTED"
+OUTCOME_NOOP = "NOOP_ALREADY_CURRENT"
+OUTCOME_STALE = "STALE_SUPERSEDED"
+OUTCOME_DIVERGENT = "REFUSED_DIVERGENT"
+OUTCOME_UNKNOWN_CURRENT = "REFUSED_UNKNOWN_CURRENT"
+
+
+# ── pure decision logic (unit-tested; no I/O) ──────────────────────────────
+
+
+def decide_promotion(candidate_sha: str, deployed_sha: str, ancestry: str) -> str:
+    """Pure promotion decision from a precomputed ancestry relationship.
+    ancestry is one of: 'same' | 'descendant' | 'ancestor' | 'divergent' |
+    'unknown'. Fail-closed: anything not provably allowed is refused."""
+    if candidate_sha == deployed_sha or ancestry == "same":
+        return OUTCOME_NOOP
+    if ancestry == "descendant":
+        return OUTCOME_PROMOTED
+    if ancestry == "ancestor":
+        return OUTCOME_STALE
+    if ancestry == "divergent":
+        return OUTCOME_DIVERGENT
+    return OUTCOME_UNKNOWN_CURRENT
+
+
+def ancestry_of(candidate_sha: str, deployed_sha: str, repo: Path, fetch_on_divergent: bool = True) -> str:
+    """Classify candidate vs deployed using git merge-base --is-ancestor on
+    the canonical release branch's repository. Returns 'unknown' on any git
+    failure (fail-closed callers treat that as refusal).
+
+    Observed live 2026-09-13 (deploy 7449, SHA 535c2909): a DELAYED older
+    candidate's checkout cannot contain commits newer than itself — its
+    local object store lacks the deployed commit entirely, so ancestry
+    classifies 'divergent' and fail-closed refuses. Correct refusal, wrong
+    reason: the candidate IS an ancestor. When the local classification is
+    divergent, fetch origin's canonical branch first (best-effort; the
+    deploy workflow's checkout has credentials configured) and re-classify.
+    If still divergent afterwards, the refusal stands with better
+    evidence."""
+
+    def _is_ancestor(a: str, b: str) -> bool:
+        r = subprocess.run(
+            ["git", "-C", str(repo), "-c", f"safe.directory={repo}", "merge-base", "--is-ancestor", a, b],
+            capture_output=True,
+            text=True,
+        )
+        return r.returncode == 0
+
+    if candidate_sha == deployed_sha:
+        return "same"
+    if _is_ancestor(candidate_sha, deployed_sha):
+        return "ancestor"  # candidate is OLDER than what's deployed
+    if _is_ancestor(deployed_sha, candidate_sha):
+        return "descendant"  # candidate is NEWER (normal forward deploy)
+    if fetch_on_divergent:
+        try:
+            subprocess.run(
+                ["git", "-C", str(repo), "-c", f"safe.directory={repo}", "fetch", "origin", "master", "--quiet"],
+                capture_output=True,
+                text=True,
+                timeout=60,
+            )
+        except (OSError, subprocess.TimeoutExpired):
+            pass  # best-effort: classification below still fails closed
+        if _is_ancestor(candidate_sha, deployed_sha):
+            return "ancestor"
+        if _is_ancestor(deployed_sha, candidate_sha):
+            return "descendant"
+    return "divergent"
+
+
+def read_deployed_commit(deploy_root: Path = DEPLOY_ROOT) -> str:
+    """Read the commit the `current` symlink ACTUALLY points to right now —
+    from the deployed build identity, the deployment authority (never from
+    a stale job-start artifact). Empty string if unreadable."""
+    current = deploy_root / "current"
+    try:
+        target = Path(os.readlink(current))
+    except OSError:
+        return ""
+    identity = target / "src" / "rune" / "_build_identity.json"
+    try:
+        data = json.loads(identity.read_text(encoding="utf-8"))
+        return str(data.get("commit", ""))
+    except (OSError, ValueError):
+        return ""
+
+
+# ── the fenced promotion operation ──────────────────────────────────────────
+
+
+def promote(
+    candidate_sha: str,
+    stage: Path,
+    tmp_stage: Path,
+    repo: Path,
+    deploy_root: Path = DEPLOY_ROOT,
+    on_promoted=None,
+) -> tuple[str, str]:
+    """Atomically decide + promote under an exclusive lock.
+
+    Returns (outcome, detail). ONLY the PROMOTED outcome mutates
+    current/last-good, and it does so inside the same critical section as
+    the decision — no interleaving window between check and swap.
+
+    Crash-safety ordering (brief §16): the stage rename happens BEFORE the
+    lock (a staged-but-unpromoted tree is inert), the symlink swap is a
+    single rename(2) via a temporary link, so current is never absent and
+    never half-updated. last-good is refreshed BEFORE the swap only from a
+    value that was genuinely current a moment ago.
+    """
+    deploy_root.mkdir(parents=True, exist_ok=True)
+    lock_path = deploy_root / ".promotion.lock"
+    with open(lock_path, "w") as lock_fd:
+        fcntl.flock(lock_fd, fcntl.LOCK_EX)
+        try:
+            deployed = read_deployed_commit(deploy_root)
+            if not deployed:
+                return (OUTCOME_UNKNOWN_CURRENT, "deployed commit unreadable — fail closed")
+            rel = ancestry_of(candidate_sha, deployed, repo)
+            outcome = decide_promotion(candidate_sha, deployed, rel)
+            if outcome != OUTCOME_PROMOTED:
+                return (outcome, f"deployed={deployed[:12]} ancestry={rel}")
+
+            # --- mutate (still inside the lock) ---
+            current = deploy_root / "current"
+            last_good = deploy_root / "last-good"
+            prev_target = os.readlink(current) if current.is_symlink() else ""
+            if prev_target and (deploy_root / "last-good").is_symlink():
+                (deploy_root / "last-good").unlink()
+            if prev_target:
+                _atomic_symlink(deploy_root, "last-good", prev_target)
+            if stage.exists():
+                # re-deploying an already-staged SHA: restage over it
+                import shutil
+
+                shutil.rmtree(stage)
+            os.rename(tmp_stage, stage)
+            _atomic_symlink_target(deploy_root, "current", stage)
+            detail = f"current -> {stage.name}"
+            if on_promoted is not None:
+                # §12C: race-safe convergence — the WINNER refreshes release
+                # history atomically with its promotion, inside the lock.
+                # Failure degrades to STALE (software valid, history lags;
+                # retry by any later deploy converges idempotently).
+                try:
+                    on_promoted(candidate_sha)
+                    detail += " (+ release-history converged)"
+                except Exception as exc:  # noqa: BLE001
+                    detail += f" (release-history refresh FAILED: {exc!s:.120} — STALE, retry converges)"
+            return (OUTCOME_PROMOTED, detail)
+        finally:
+            fcntl.flock(lock_fd, fcntl.LOCK_UN)
+
+
+def _atomic_symlink(deploy_root: Path, name: str, target: str) -> None:
+    """Symlink swap that never leaves `name` absent: write temp link, rename
+    over the final name (rename(2) is atomic on the same filesystem)."""
+    _atomic_symlink_target(deploy_root, name, Path(target))
+
+
+def _atomic_symlink_target(deploy_root: Path, name: str, target: Path) -> None:
+    tmp = deploy_root / f".{name}.swap-{os.getpid()}"
+    if tmp.is_symlink() or tmp.exists():
+        tmp.unlink()
+    tmp.symlink_to(target)
+    os.replace(tmp, deploy_root / name)
@@ -6,6 +6,16 @@ has no embedded multi-line scripts — a YAML block-scalar indentation hazard).

 Prints one of: success | failure | pending
 Exit code 0 always (the shell decides what to do with the verdict).
+
+RELEASE-SAFETY HARDENING (2026-09-13, after red-CI promotions 0.66.28/29):
+the verdict is the AND of two independent legs —
+  1. required-checks context (newest observation for this exact SHA);
+  2. the latest COMPLETED ci.yml@refs/heads/master run for this exact SHA,
+     every job concluded success (branch+tag fan-out resolved by path;
+     jobs envelope unwrapped — no aggregate-commit-status fallback, the
+     exact bug class verify.py hit in W/R-v0.3).
+Any disagreement, pending, or missing evidence → pending/failure.
+FAIL-CLOSED: not verified → not deployable.
 """

 from __future__ import annotations
@@ -14,38 +24,152 @@ import json
 import os
 import sys
 import urllib.request
+import sys as _sys
+_SCRIPT_DIR = str(__import__('pathlib').Path(__file__).resolve().parent)
+if _SCRIPT_DIR not in _sys.path:
+    _sys.path.insert(0, _SCRIPT_DIR)
+import gitea_auth

+REQUIRED_CONTEXT = "CI Pipeline / required-checks"
+CI_WORKFLOW_PATH = "ci.yml@refs/heads/master"
+
+
+# ── pure verdict logic (unit-tested; fixtures from the real incidents) ────
+
+def required_check_verdict(statuses: list[dict]) -> str:
+    """Newest non-skipped required-checks observation for the SHA.
+    Missing → pending. Newest wins (stale success never overrides a newer
+    failure, and vice versa)."""
+    candidates = [
+        s for s in statuses
+        if str(s.get("context", "")).startswith(REQUIRED_CONTEXT)
+        and s.get("status") != "skipped"
+    ]
+    if not candidates:
+        return "pending"
+    newest = max(candidates, key=lambda s: str(s.get("created_at", "")))
+    return str(newest.get("status", "pending"))
+
+
+def independent_jobs_verdict(runs: list[dict], jobs_fetch, sha: str = "") -> str:
+    """Latest COMPLETED ci.yml@refs/heads/master run for the exact SHA:
+    every job must conclude success. Fan-out-safe (tag runs and sibling
+    workflows are excluded by path); the run is resolved by head_sha, never
+    by 'latest branch state'. Missing run → pending. Any non-success
+    conclusion → failure."""
+    mine = [r for r in runs if r.get("path") == CI_WORKFLOW_PATH]
+    if sha:
+        mine = [r for r in mine if str(r.get("head_sha", "")) == sha]
+    completed = [r for r in mine if r.get("status") == "completed"]
+    if not completed:
+        return "pending"
+    latest = max(completed, key=lambda r: str(r.get("created_at", "") or r.get("run_started_at", "")))
+    jobs = jobs_fetch(latest.get("id"))
+    if jobs is None:
+        return "pending"  # jobs probe failed → not verified
+    if not jobs:
+        return "pending"  # no jobs visible → cannot verify
+    for j in jobs:
+        if j.get("conclusion") != "success":
+            return "failure"
+    return "success"
+
+
+def combine(a: str, b: str) -> str:
+    """Fail-closed conjunction: any failure → failure; any pending → pending;
+    success only when BOTH legs are success."""
+    if "failure" in (a, b):
+        return "failure"
+    if "pending" in (a, b):
+        return "pending"
+    return "success"
+
+
+def qualify(statuses: list[dict], runs: list[dict], jobs_fetch, sha: str = "") -> str:
+    return combine(required_check_verdict(statuses),
+                   independent_jobs_verdict(runs, jobs_fetch, sha=sha))
+
+
+# ── API fetches (thin; injectable in tests) ───────────────────────────────
+
+def _get_json(url: str, token: str, timeout: int = 30):
+    req = urllib.request.Request(url, headers={"Authorization": f"token {token}"})
+    try:
+        with urllib.request.urlopen(req, timeout=timeout) as r:
+            return json.loads(r.read())
+    except (OSError, ValueError):
+        return None
+
+
+def _fetch_status(server: str, repo: str, sha: str, token: str) -> list[dict] | None:
+    data = _get_json(f"{server}/api/v1/repos/{repo}/commits/{sha}/status", token)
+    return None if data is None else data.get("statuses", [])
+
+
+def _fetch_runs(server: str, repo: str, sha: str, token: str,
+                max_pages: int = 4) -> list[dict] | None:
+    """H-PATCH: paginate the runs list. The repo's scheduled merge-robot
+    runs flood the first page (~50 slots exhaust within hours), pushing
+    older-but-relevant SHAs to later pages — the cause of the observed
+    authority flapping (terminal commit read as pending). Paginate until
+    the sha is found or max_pages exhausted."""
+    out: list[dict] = []
+    for page in range(1, max_pages + 1):
+        data = _get_json(
+            f"{server}/api/v1/repos/{repo}/actions/runs?limit=50&page={page}",
+            token)
+        if data is None:
+            return None if page == 1 else (out or None)
+        runs = data.get("workflow_runs", data if isinstance(data, list) else [])
+        out.extend(r for r in runs if str(r.get("head_sha", "")) == sha)
+        if len(runs) < 50:
+            break  # exhausted
+    return out
+
+
+def _fetch_jobs(server: str, repo: str, run_id, token: str) -> list[dict] | None:
+    data = _get_json(f"{server}/api/v1/repos/{repo}/actions/runs/{run_id}/jobs", token)
+    if data is None:
+        return None
+    # jobs envelope unwrap (this Gitea wraps in {"jobs": [...]}; some
+    # versions return a bare list) — never fall back to aggregate status
+    if isinstance(data, dict):
+        return data.get("jobs", [])
+    return data if isinstance(data, list) else None
+
+
+# ── CLI ───────────────────────────────────────────────────────────────────

 def main() -> int:
-    server = os.environ.get("GITEA_SERVER", "").rstrip("/")
-    repo = os.environ.get("GITEA_REPO", "")
-    sha = os.environ.get("DEPLOY_SHA", "")
-    token = os.environ.get("GITEA_TOKEN", "")
+    # T21.4 fix: fall back to the canonical defaults when the caller did
+    # not export GITEA_SERVER/GITEA_REPO (CI always exports them; local
+    # and first-class-tool consumers previously got a silent "pending"
+    # verdict — a wrong answer, not a fail-closed one).
+    # H-PATCH H4: the target sha arrives EXPLICITLY (--sha) from callers
+    # that hold structured operation state. DEPLOY_SHA env remains only
+    # for standalone/workflow backward compat — the canonical path never
+    # depends on ambient env for operation identity.
+    ap = __import__("argparse").ArgumentParser()
+    ap.add_argument("--sha", dest="sha_arg", default=None)
+    args, _unknown = ap.parse_known_args()
+    server = (os.environ.get("GITEA_SERVER", "")
+              or gitea_auth.DEFAULT_SERVER).rstrip("/")
+    repo = os.environ.get("GITEA_REPO", "") or gitea_auth.DEFAULT_REPO
+    sha = args.sha_arg or os.environ.get("DEPLOY_SHA", "")
+    token = gitea_auth.get_token() if gitea_auth.has_token() else ""
    if not (server and repo and sha and token):
        print("pending")  # missing config → cannot verify → not qualified
        return 0
-    url = f"{server}/api/v1/repos/{repo}/commits/{sha}/status"
-    req = urllib.request.Request(url, headers={"Authorization": f"token {token}"})
-    try:
-        with urllib.request.urlopen(req, timeout=30) as r:
-            data = json.loads(r.read())
-    except (OSError, ValueError) as exc:
-        # transient API/network failure → keep polling (urllib raises
-        # URLError (OSError subclass); bad JSON raises ValueError)
-        print(f"pending (status probe failed: {exc})")
+
+    statuses = _fetch_status(server, repo, sha, token)
+    runs = _fetch_runs(server, repo, sha, token)
+    if statuses is None or runs is None:
+        # transient API/network failure → keep polling
+        print("pending (probe failed)")
        return 0
-    required = [
-        s
-        for s in data.get("statuses", [])
-        if (
-            str(s.get("context", "")).startswith("CI Pipeline / required-checks")
-            and s.get("status") != "skipped"
-        )
-    ]
-    if not required:
-        print("pending")
-    else:
-        print(required[0].get("status", "pending"))
+    jobs_fetch = lambda run_id: _fetch_jobs(server, repo, run_id, token)  # noqa: E731
+    verdict = qualify(statuses, runs, jobs_fetch, sha=sha)
+    print(verdict)
    return 0


@@ -3,12 +3,15 @@

 Called by .gitea/workflows/deploy.yml. Stages src/ into
 /mnt/workdrive/rune-code/<short-sha>, stamps _build_identity.json, verifies
-the staged tree (import + SHA assert + authority smoke), then ATOMICALLY
-promotes the `current` symlink and records `last-good` (safe rollback:
-the previous deployment remains available).
+the staged tree (import + SHA assert + authority smoke), then promotes
+through the MONOTONIC promotion authority (deploy_promotion.py): an older
+candidate finishing after a newer promoted one is fenced out (STALE_SUPERSEDED)
+and reports truthfully — build/stage success != promotion success.

-Exit 0 = deployed and verified. Non-zero = deployment refused; the previous
-deployment stays current.
+Exit 0 = PROMOTED and verified, or NOOP_ALREADY_CURRENT, or a truthful
+STALE_SUPERSEDED (newer build already current; nothing mutated).
+Non-zero = deployment refused (divergent/unknown ordering, verification
+failure); the previous deployment stays current.
 """

 from __future__ import annotations
@@ -25,6 +28,92 @@ DEPLOY_ROOT = Path("/mnt/workdrive/rune-code")
 REPO_ROOT = Path(__file__).resolve().parent.parent
 VENV_PY = "/home/nick/.venvs/memory/bin/python3"

+sys.path.insert(0, str(REPO_ROOT / "scripts"))
+sys.path.insert(0, str(REPO_ROOT / "src"))
+import deploy_promotion  # noqa: E402
+from rune.release_notes import (  # noqa: E402
+    ReleaseNote as _ReleaseNote,
+    auto_initial_note as _auto_note,
+    converged_deployed_sha as _authority_sha,
+    convergence_refresh as _refresh,
+    fetch_release as _fetch_release,
+    publish_release as _publish,
+    sha_from_body as _sha_from_body,
+    summary_from_body as _summary_from_body,
+)
+
+
+def publish_out_of_lock(short: str, sha: str, *,
+                        deploy_root: Path = DEPLOY_ROOT,
+                        scaffold_store: Path | None = None) -> None:
+    """§13D/§14B: release-history publication AFTER the promotion lock is
+    released. Network I/O never holds the flock.
+
+    §14B INVARIANT: publication does NOT trust the originating job's SHA.
+    The CURRENT deployment authority is re-read immediately before
+    publishing; the note is constructed from CURRENT authority. A stale
+    job whose candidate was superseded converges the NEW authority (or
+    refuses a version it cannot truthfully describe) — it may help
+    converge B, but may never reassert itself as deployed.
+    Failure degrades to STALE (software valid, history lags; retry
+    converges)."""
+    try:
+        server = os.environ.get("GITEA_SERVER", "").rstrip("/")
+        repo = os.environ.get("GITEA_REPO", "")
+        token = os.environ.get("GITEA_TOKEN", "")
+        if not (server and repo and token):
+            return
+        # §14B: re-read CURRENT authority — never trust the job's SHA.
+        authority_sha = _authority_sha(str(deploy_root))
+        if not authority_sha:
+            print("WARNING: release-history convergence skipped: "
+                  "deployment authority unreadable — STALE, retry converges")
+            return
+        store = str(scaffold_store or (REPO_ROOT / "store" / "scaffold" / "current.json"))
+        identity = deploy_root / authority_sha[:12] / "src" / "rune" / "_build_identity.json"
+        try:
+            version = json.loads(identity.read_text()).get("version", "")
+        except (OSError, ValueError):
+            version = ""
+        if not version:
+            print("WARNING: release-history convergence skipped: "
+                  "no version on current authority — STALE, retry converges")
+            return
+        tag = f"v{version}"
+        existing = _fetch_release(server, repo, token, tag)
+        if existing is None:
+            note = _auto_note(
+                version=version,
+                summary=(f"Rune v{version} — evidence-derived release "
+                         f"history (automatic convergence)"),
+                scaffold_store=store,
+                release_sha=authority_sha, deployed_sha=authority_sha,
+                provenance=["automatic convergence publish "
+                            "(deploy workflow, authority-derived)"])
+        else:
+            note = _refresh(
+                _ReleaseNote.from_dict({
+                    "version": version,
+                    "summary": _summary_from_body(existing.get("body", "")),
+                    "release_sha": _sha_from_body(existing.get("body", ""), "tag")
+                    or authority_sha,
+                    "provenance": [],
+                }),
+                deployed_sha=authority_sha, scaffold_store=store,
+                extra_provenance="automatic convergence publish "
+                                "(deploy workflow, authority-derived)")
+            # preserve the tag SHA the stored body carries
+            tag_sha = _sha_from_body(existing.get("body", ""), "tag")
+            if tag_sha:
+                note.release_sha = tag_sha
+        result = _publish(server, repo, token, note)
+        print(f"release-history convergence: "
+              f"{result.get('action', 'unknown')} "
+              f"(authority {authority_sha[:12]})")
+    except Exception as conv_exc:  # noqa: BLE001 — history STALE, software valid
+        print(f"WARNING: release-history convergence failed: "
+              f"{conv_exc!s:.200} — STALE, retry converges")
+

 def main() -> int:
    # The runner's checkout workspace (act sets GITHUB_WORKSPACE). Never a
@@ -99,7 +188,14 @@ print("VERIFIED:", json.dumps(info))
    # RuneApp initialization must succeed before promotion. A staged
    # revision whose TUI cannot initialize never becomes current.
    canary = subprocess.run(
-        [VENV_PY, str(REPO_ROOT / "scripts" / "tui_startup_canary.py")],
+        [
+            VENV_PY,
+            str(REPO_ROOT / "scripts" / "tui_startup_canary.py"),
+            "--mode",
+            "candidate",
+            "--stage",
+            str(tmp_stage),
+        ],
        capture_output=True,
        text=True,
        env={**env, "PYTHONPATH": str(tmp_stage / "src")},
@@ -112,25 +208,49 @@ print("VERIFIED:", json.dumps(info))
        shutil.rmtree(tmp_stage, ignore_errors=True)
        return 1

-    # --- promote (atomic) + last-known-good ------------------------------
-    DEPLOY_ROOT.mkdir(parents=True, exist_ok=True)
-    current = DEPLOY_ROOT / "current"
-    if current.is_symlink():
-        last_good = DEPLOY_ROOT / "last-good"
-        if last_good.is_symlink():
-            last_good.unlink()
-        last_good.symlink_to(os.readlink(current))
-    if current.is_symlink() or current.exists():
-        current.unlink()
-    if stage.exists():
-        shutil.rmtree(stage)  # re-deploying an already-staged SHA: restage
-    os.rename(tmp_stage, stage)
-    current.symlink_to(stage)
-    print(f"PROMOTED: {short}")
-    print("current ->", os.readlink(current))
-    if (DEPLOY_ROOT / "last-good").is_symlink():
-        print("last-good ->", os.readlink(DEPLOY_ROOT / "last-good"))
-    return 0
+    # --- promote: FENCED through the promotion authority -----------------
+    # SCAFFOLD contract DEPLOY-MONOTONIC-001: promotion is monotonic. The
+    # ancestry check + lock live in deploy_promotion.py; this script only
+    # handles the truthful outcome reporting. A stale candidate must NOT
+    # claim promotion success — build/stage success != promotion success.
+    #
+    # §14C: NO separate convergence-intent ledger. The durable pending-
+    # convergence state IS the deployment authority itself (atomically
+    # swapped `current` symlink + stamped build identity): publication
+    # derives from it, idempotently, outside the lock. The old intent
+    # file had exactly one writer and zero readers, non-atomic writes,
+    # and a bogus recorded_at (SECONDS is a bash builtin, never exported
+    # — always "0"). Deriving from authority gives atomicity, restart
+    # readability, corruption immunity, monotonicity (stale publishers
+    # read CURRENT authority, not their own job) and idempotent retry
+    # for free. REMOVE more complexity than you add.
+
+    outcome, detail = deploy_promotion.promote(
+        sha, DEPLOY_ROOT / short, tmp_stage, repo)
+
+    # §13D: publish release history OUTSIDE the promotion lock (network
+    # I/O must not hold the flock). Only for PROMOTED outcomes.
+    if outcome == deploy_promotion.OUTCOME_PROMOTED:
+        publish_out_of_lock(short, sha)
+
+    if outcome == deploy_promotion.OUTCOME_PROMOTED:
+        print(f"PROMOTED: {short}")
+        print("current ->", os.readlink(DEPLOY_ROOT / "current"))
+        if (DEPLOY_ROOT / "last-good").is_symlink():
+            print("last-good ->", os.readlink(DEPLOY_ROOT / "last-good"))
+        return 0
+    if outcome == deploy_promotion.OUTCOME_NOOP:
+        print(f"NOOP_ALREADY_CURRENT: {short} is the deployed commit")
+        print(detail)
+        return 0
+    if outcome == deploy_promotion.OUTCOME_STALE:
+        print(f"STALE_SUPERSEDED: {short} is older than the deployed commit — promotion refused (monotonic promotion)")
+        print(detail)
+        print("current unchanged:", os.readlink(DEPLOY_ROOT / "current"))
+        return 0  # truthful non-error terminal state: the newer build won
+    # divergent / unknown-current: fail closed
+    print(f"REFUSED ({outcome}): {short} — {detail}", file=sys.stderr)
+    return 1


 if __name__ == "__main__":
@@ -12,6 +12,7 @@ Canonical pipeline (RP2, handoff 2026-09-08):
 Per the authoritative build prompt §18:
  "DETERMINISTIC TESTS PROVE THE PIPE. REAL MIMO DOES THE RESEARCH."
 """
+
 from __future__ import annotations

 import json
@@ -21,6 +22,7 @@ from pathlib import Path
 sys.path.insert(0, str(Path(__file__).parent))
 from research_task_tree import ResearchTaskTree, STAGES

+
 def fake_issue_gate(tree: ResearchTaskTree) -> dict:
    """ISSUE is deterministic: the Gitea issue content is the DEFINE replacement.
    The fake worker emits the compiled gate artifact directly."""
@@ -33,6 +35,7 @@ def fake_issue_gate(tree: ResearchTaskTree) -> dict:
        "quality": "BUILDABLE",
    }

+
 def fake_inward(tree: ResearchTaskTree) -> dict:
    # Assert ISSUE-gate artifact was actually supplied
    prior = tree.get_prior_artifacts("INWARD")
@@ -40,11 +43,22 @@ def fake_inward(tree: ResearchTaskTree) -> dict:
    assert issue_gate.get("goal"), "INWARD: ISSUE-gate artifact missing — evidence handoff broken"
    return {
        "claims": [
-            {"id": "I1", "claim": "memory_store uses FTS5", "evidence": ["src/rune/memory_store.py:86"], "status": "CURRENT"},
-            {"id": "I2", "claim": "shard routing is present", "evidence": ["src/rune/shard_manager.py:114"], "status": "CURRENT"},
+            {
+                "id": "I1",
+                "claim": "memory_store uses FTS5",
+                "evidence": ["src/rune/memory_store.py:86"],
+                "status": "CURRENT",
+            },
+            {
+                "id": "I2",
+                "claim": "shard routing is present",
+                "evidence": ["src/rune/shard_manager.py:114"],
+                "status": "CURRENT",
+            },
        ],
    }

+
 def fake_outward(tree: ResearchTaskTree) -> dict:
    """Unified OUTWARD (RP3): one research task, discovery + reading as
    SUBSECTIONS of one artifact — not separate stages."""
@@ -58,19 +72,35 @@ def fake_outward(tree: ResearchTaskTree) -> dict:
            "tool_receipts": [{"tool": "webfetch", "query": "fts5 hybrid search", "status": "completed"}],
            "sources": [
                {"id": "O1", "url": "https://sqlite.org/fts5.html", "title": "FTS5", "source_class": "software"},
-                {"id": "O2", "url": "https://github.com/asg017/sqlite-vec", "title": "sqlite-vec", "source_class": "software"},
+                {
+                    "id": "O2",
+                    "url": "https://github.com/asg017/sqlite-vec",
+                    "title": "sqlite-vec",
+                    "source_class": "software",
+                },
            ],
        },
        "reading": {
            "evidence": [
-                {"id": "E1", "source_id": "O1", "url": "https://sqlite.org/fts5.html",
-                 "claim": "FTS5 supports external content tables", "limitations": ["token porter only"]},
-                {"id": "E2", "source_id": "O2", "url": "https://github.com/asg017/sqlite-vec",
-                 "claim": "vec0 virtual tables for embedding search", "limitations": []},
+                {
+                    "id": "E1",
+                    "source_id": "O1",
+                    "url": "https://sqlite.org/fts5.html",
+                    "claim": "FTS5 supports external content tables",
+                    "limitations": ["token porter only"],
+                },
+                {
+                    "id": "E2",
+                    "source_id": "O2",
+                    "url": "https://github.com/asg017/sqlite-vec",
+                    "claim": "vec0 virtual tables for embedding search",
+                    "limitations": [],
+                },
            ],
        },
    }

+
 def fake_applicability(tree: ResearchTaskTree) -> dict:
    """APPLICABILITY (RP4): classify researched mechanisms against Rune's reality."""
    prior = tree.get_prior_artifacts("APPLICABILITY")
@@ -96,6 +126,7 @@ def fake_applicability(tree: ResearchTaskTree) -> dict:
        ],
    }

+
 def fake_adversarial(tree: ResearchTaskTree) -> dict:
    prior = tree.get_prior_artifacts("ADVERSARIAL")
    outward = prior.get("OUTWARD", {})
@@ -104,19 +135,31 @@ def fake_adversarial(tree: ResearchTaskTree) -> dict:
    return {
        "disposition": "SURVIVES_WITH_MITIGATIONS",
        "counter_evidence": [
-            {"id": "A1", "source_id": "O1", "claim": "FTS5 porter tokenizer may not handle all languages",
-             "derived_from": ["E1"], "limitations": ["minor"]},
+            {
+                "id": "A1",
+                "source_id": "O1",
+                "claim": "FTS5 porter tokenizer may not handle all languages",
+                "derived_from": ["E1"],
+                "limitations": ["minor"],
+            },
        ],
        "surviving_candidates": ["M1"],
        "rejected_candidates": [],
        "required_mitigations": ["mode labeling"],
        "falsification_tests": ["FT1: harness run with embedding backend unreachable"],
        "attacks": [
-            {"id": "A1", "target_candidate": "M1", "dimension": "failure modes",
-             "finding": "tokenizer coverage", "evidence": ["E1"], "severity": "minor"},
+            {
+                "id": "A1",
+                "target_candidate": "M1",
+                "dimension": "failure modes",
+                "finding": "tokenizer coverage",
+                "evidence": ["E1"],
+                "severity": "minor",
+            },
        ],
    }

+
 def fake_innovation(tree: ResearchTaskTree) -> dict:
    prior = tree.get_prior_artifacts("INNOVATION")
    inward = prior.get("INWARD", {})
@@ -131,14 +174,21 @@ def fake_innovation(tree: ResearchTaskTree) -> dict:
    return {
        "conclusion": "ADOPT",
        "candidates": [
-            {"id": "C1", "category": "ADAPT", "proposal": "hybrid FTS+vec recall",
-             "derived_from": inward_claims[:1] + evidence_ids[:1] + applicable[:1],
-             "composition_of": [], "novelty": "adapted",
-             "expected_gain": "better recall", "risks": ["latency"],
-             "kill_test": "recall returns >0 rows"},
+            {
+                "id": "C1",
+                "category": "ADAPT",
+                "proposal": "hybrid FTS+vec recall",
+                "derived_from": inward_claims[:1] + evidence_ids[:1] + applicable[:1],
+                "composition_of": [],
+                "novelty": "adapted",
+                "expected_gain": "better recall",
+                "risks": ["latency"],
+                "kill_test": "recall returns >0 rows",
+            },
        ],
    }

+
 def fake_synthesis(tree: ResearchTaskTree) -> dict:
    prior = tree.get_prior_artifacts("SYNTHESIS")
    innovation = prior.get("INNOVATION", {})
@@ -157,6 +207,7 @@ def fake_synthesis(tree: ResearchTaskTree) -> dict:
        "build_instructions": ["modify recall() in memory_store.py", "add tests"],
    }

+
 def fake_qualify(tree: ResearchTaskTree) -> dict:
    prior = tree.get_prior_artifacts("QUALIFY")
    synthesis = prior.get("SYNTHESIS", {})
@@ -177,6 +228,7 @@ def fake_qualify(tree: ResearchTaskTree) -> dict:
        "evidence_used": synthesis.get("evidence_used", []),
    }

+
 FAKE_WORKERS = {
    "ISSUE": fake_issue_gate,
    "INWARD": fake_inward,
@@ -188,6 +240,7 @@ FAKE_WORKERS = {
    "QUALIFY": fake_qualify,
 }

+
 def run_fake_stage(tree: ResearchTaskTree) -> dict:
    stage = tree.current_stage
    worker = FAKE_WORKERS.get(stage)
@@ -195,8 +248,10 @@ def run_fake_stage(tree: ResearchTaskTree) -> dict:
        raise ValueError(f"no fake worker for stage {stage}")
    return worker(tree)

+
 def main_with(argv: list[str]) -> int:
    import argparse
+
    parser = argparse.ArgumentParser(description="Fake stage worker for pipeline testing")
    parser.add_argument("--issue", type=int, required=True)
    parser.add_argument("--round", type=int, default=1)
@@ -216,5 +271,6 @@ def main_with(argv: list[str]) -> int:
    print(f"fake stage {stage} complete")
    return 0

+
 if __name__ == "__main__":
    raise SystemExit(main_with(sys.argv))
@@ -20,6 +20,11 @@ import urllib.error
 import urllib.parse
 import urllib.request
 from typing import Any
+import sys as _sys
+_SCRIPT_DIR = str(__import__('pathlib').Path(__file__).resolve().parent)
+if _SCRIPT_DIR not in _sys.path:
+    _sys.path.insert(0, _SCRIPT_DIR)
+import gitea_auth

 DRIFT_ISSUE_LABEL = "release-drift"
 DRIFT_ISSUE_LABEL_COLOR = "ff0000"
@@ -108,7 +113,7 @@ def main(argv: list[str] | None = None, api: Any = None) -> int:
        print(body)
        return 0

-    token = os.environ["GITEA_TOKEN"]
+    token = gitea_auth.get_token()
    base = os.environ["RELEASE_SERVER_URL"] + "/api/v1/repos/" + os.environ["RELEASE_REPO"]

    if api is None:
@@ -0,0 +1,112 @@
+"""gitea_auth.py — canonical Gitea API authentication (handoff 2026-09-11 §H).
+
+One deterministic helper for authenticated Gitea API access. Never prints
+credentials. Fails clearly when unavailable. Credential ladder:
+
+  1. GITEA_TOKEN environment variable (preferred for CI/deployment)
+  2. private token file (GITEA_TOKEN_FILE, default
+     ~/.config/opencode/secrets/gitea-token, must be mode 0600)
+  3. git credential fill (uses the credential helper / store)
+
+Never parses credentials from git remote URLs.
+"""
+
+from __future__ import annotations
+
+import json
+import os
+import stat
+import subprocess
+import urllib.request
+from pathlib import Path
+from typing import Any
+
+DEFAULT_SERVER = "https://gitea.thingscouldgetdicey.online"
+DEFAULT_REPO = "conductor/rune"
+DEFAULT_TOKEN_FILE = Path("~/.config/opencode/secrets/gitea-token").expanduser()
+GITEA_HOST = "gitea.thingscouldgetdicey.online"
+
+
+class GiteaCredentialError(RuntimeError):
+    """Raised when no Gitea credential is available. Message never contains one."""
+
+
+def _token_from_file() -> str:
+    path = Path(os.environ.get("GITEA_TOKEN_FILE", DEFAULT_TOKEN_FILE)).expanduser()
+    if not path.is_file():
+        return ""
+    mode = stat.S_IMODE(path.stat().st_mode)
+    if mode & 0o077:
+        raise PermissionError(f"Gitea token file must be mode 0600: {path}")
+    return path.read_text(encoding="utf-8").strip()
+
+
+def _token_from_credential_helper() -> str:
+    try:
+        result = subprocess.run(
+            ["git", "credential", "fill"],
+            input=f"protocol=https\nhost={GITEA_HOST}\nusername=conductor\n\n",
+            capture_output=True,
+            text=True,
+            timeout=10,
+        )
+        for line in result.stdout.splitlines():
+            if line.startswith("password="):
+                return line[len("password=") :]
+    except (subprocess.TimeoutExpired, OSError):
+        pass
+    return ""
+
+
+def has_token() -> bool:
+    """True if any rung of the credential ladder answers. Never raises."""
+    try:
+        get_token()
+        return True
+    except Exception:  # noqa: BLE001 — probe-only: any failure means "no credential available"
+        return False
+
+
+def get_token() -> str:
+    """Obtain the Gitea API token. Never prints the token."""
+    token = os.environ.get("GITEA_TOKEN", "").strip()
+    if token:
+        return token
+    token = _token_from_file()
+    if token:
+        return token
+    token = _token_from_credential_helper()
+    if token:
+        return token
+    raise GiteaCredentialError(
+        "Gitea credentials unavailable: set GITEA_TOKEN, create a mode-0600 "
+        f"token file at {DEFAULT_TOKEN_FILE}, or configure a git credential "
+        f"helper for {GITEA_HOST}"
+    )
+
+
+def api(
+    path: str,
+    method: str = "GET",
+    data: dict | None = None,
+    server: str = DEFAULT_SERVER,
+    repo: str = DEFAULT_REPO,
+) -> tuple[int, Any]:
+    """Make an authenticated Gitea API request. Returns (status_code, parsed_body)."""
+    token = get_token()
+    url = f"{server}/api/v1/repos/{repo}{path}"
+    req = urllib.request.Request(
+        url,
+        data=json.dumps(data).encode() if data is not None else None,
+        headers={"Authorization": f"token {token}", "Content-Type": "application/json"},
+        method=method,
+    )
+    try:
+        with urllib.request.urlopen(req, timeout=30) as r:
+            body, status = r.read(), r.status
+    except urllib.error.HTTPError as e:
+        body, status = e.read(), e.code
+    try:
+        return status, (json.loads(body) if body else {})
+    except json.JSONDecodeError:
+        return status, {"_raw": body.decode("utf-8", "replace")[:200]}
@@ -0,0 +1,110 @@
+#!/usr/bin/env python3
+"""gitea_reconciliation_check.py — Part L consistency check (deterministic).
+
+Compares live Gitea issue state against the canonical reconciliation
+artifact (docs/gitea_issue_reconciliation.json). Flags:
+
+  1. closed issue recorded in an open-only state
+     (VALID_OPEN / DEFERRED_FUTURE / NEEDS_REVALIDATION)
+  2. open issue recorded in a terminal state
+     (FIXED_VERIFIED / SUPERSEDED / DUPLICATE / HISTORICAL_EVIDENCE /
+     ACCEPTED_LIMITATION)
+  3. `bug` label present on any issue not recorded live_defect
+     (or missing on a recorded live defect)
+  4. open issue missing from the artifact
+  5. record missing required fields
+
+Exit 0 = consistent. Exit 1 = drift (each item printed). Read-only.
+"""
+from __future__ import annotations
+
+import json
+import sys
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parent.parent
+sys.path.insert(0, str(ROOT / "scripts"))
+
+ARTIFACT = ROOT / "docs" / "gitea_issue_reconciliation.json"
+
+OPEN_ONLY_STATES = {"VALID_OPEN", "DEFERRED_FUTURE", "NEEDS_REVALIDATION"}
+TERMINAL_STATES = {"FIXED_VERIFIED", "SUPERSEDED", "DUPLICATE", "HISTORICAL_EVIDENCE", "ACCEPTED_LIMITATION"}
+REQUIRED_FIELDS = {"number", "tracker_state", "reconciliation_state", "labels", "live_defect", "last_revalidated"}
+
+
+def main() -> int:
+    import gitea_auth
+
+    if not ARTIFACT.is_file():
+        print(f"reconciliation artifact missing: {ARTIFACT}")
+        return 1
+    artifact = json.loads(ARTIFACT.read_text())
+    records = {r["number"]: r for r in artifact["records"]}
+
+    drift: list[str] = []
+
+    # Fetch live issues (open + recently closed via both pages)
+    issues: list[dict] = []
+    for page in (1, 2):
+        status, batch = gitea_auth.api(f"/issues?state=all&limit=50&page={page}")
+        if status != 200:
+            print(f"Gitea API failed ({status}): {str(batch)[:200]}")
+            return 1
+        if isinstance(batch, list):
+            issues.extend(batch)
+    live = {i["number"]: i for i in issues}
+
+    for num, rec in sorted(records.items()):
+        for field in REQUIRED_FIELDS:
+            if field not in rec:
+                drift.append(f"#{num}: record missing required field {field!r}")
+        state = rec.get("reconciliation_state")
+        if state in OPEN_ONLY_STATES and rec.get("tracker_state") == "closed":
+            drift.append(f"#{num}: closed issue recorded {state}")
+        if state in TERMINAL_STATES and rec.get("tracker_state") == "open":
+            drift.append(f"#{num}: open issue recorded terminal {state}")
+        has_bug = "bug" in (rec.get("labels") or [])
+        # The bug-label/live-defect invariant applies to OPEN issues only:
+        # a closed issue's labels are historical record; its reconciliation
+        # state governs (FIXED_VERIFIED issues legitimately keep the label).
+        if rec.get("tracker_state") == "open":
+            if has_bug and not rec.get("live_defect"):
+                drift.append(f"#{num}: bug label present but not recorded as live defect")
+            if not has_bug and rec.get("live_defect"):
+                drift.append(f"#{num}: recorded live defect but bug label missing")
+        if state == "FIXED_VERIFIED" and rec.get("tracker_state") == "open":
+            drift.append(f"#{num}: FIXED_VERIFIED but still open without closure note")
+
+    # Live vs artifact sync (only for issues present in the live window)
+    for num, li in sorted(live.items()):
+        if num not in records:
+            continue
+        rec = records[num]
+        live_state = li["state"]
+        if live_state == "closed" and rec["tracker_state"] == "open" and rec["reconciliation_state"] in OPEN_ONLY_STATES:
+            drift.append(f"#{num}: live state closed but artifact says open/{rec['reconciliation_state']}")
+        if live_state == "open" and rec["tracker_state"] == "closed" and rec["reconciliation_state"] in TERMINAL_STATES:
+            drift.append(f"#{num}: live state open but artifact says closed/{rec['reconciliation_state']}")
+        live_bug = "bug" in [l["name"] for l in li.get("labels", [])]
+        if live_bug and not rec.get("live_defect"):
+            drift.append(f"#{num}: live bug label present but artifact records no live defect")
+        if not live_bug and rec.get("live_defect"):
+            drift.append(f"#{num}: artifact records live defect but live bug label absent")
+
+    # Live defects in the artifact must all be OPEN in the tracker
+    for num in artifact.get("live_defects", []):
+        li = live.get(num)
+        if li is not None and li["state"] != "open":
+            drift.append(f"#{num}: live defect but tracker state is {li['state']} — close the defect or update the artifact")
+
+    if drift:
+        print(f"RECONCILIATION DRIFT ({len(drift)}):")
+        for d in drift:
+            print(f"  - {d}")
+        return 1
+    print(f"reconciliation consistent ({len(records)} records, {len(live)} live issues in window)")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
@@ -0,0 +1,44 @@
+#!/usr/bin/env python3
+"""install_mimo_hook.py — T21.2 §21.2D: canonical MiMo hook installer.
+
+Canonical source: config/mimo-hooks/event-wait-guard.ts
+Target: ~/.mimocode/hooks/event-wait-guard.ts
+Verifies hash after install. Detects drift in both directions."""
+import hashlib
+import shutil
+import sys
+from pathlib import Path
+
+CANONICAL = Path(__file__).resolve().parent.parent / "config" / "mimo-hooks" / "event-wait-guard.ts"
+TARGET = Path.home() / ".mimocode" / "hooks" / "event-wait-guard.ts"
+
+
+def sha256(p: Path) -> str:
+    return hashlib.sha256(p.read_bytes()).hexdigest()
+
+
+def main() -> int:
+    if not CANONICAL.is_file():
+        print(f"canonical source missing: {CANONICAL}")
+        return 1
+    canon_hash = sha256(CANONICAL)
+    if TARGET.is_file():
+        target_hash = sha256(TARGET)
+        if target_hash == canon_hash:
+            print(f"in-sync: {target_hash[:12]}")
+            return 0
+        print(f"DRIFT: canonical={canon_hash[:12]} installed={target_hash[:12]}")
+    else:
+        print(f"hook missing at {TARGET}")
+    TARGET.parent.mkdir(parents=True, exist_ok=True)
+    shutil.copy2(CANONICAL, TARGET)
+    installed_hash = sha256(TARGET)
+    if installed_hash != canon_hash:
+        print(f"INSTALL FAILED: hash mismatch after copy")
+        return 1
+    print(f"installed: {installed_hash[:12]}")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
@@ -0,0 +1,52 @@
+#!/usr/bin/env python3
+"""install_mimo_tools.py — canonical MiMo tool installer (TTY-repair tranche).
+
+Canonical sources: config/mimo-tools/{bash.ts,await-operation.ts}
+Targets:           ~/.mimocode/tools/{bash.ts,await-operation.ts}
+Verifies hashes after install; detects drift in both directions
+(the parity test reads these hashes).
+"""
+import hashlib
+import shutil
+import sys
+from pathlib import Path
+
+REPO = Path(__file__).resolve().parent.parent
+TOOLS = ["bash.ts", "await-operation.ts"]
+
+
+def sha256(p: Path) -> str:
+    return hashlib.sha256(p.read_bytes()).hexdigest()
+
+
+def main() -> int:
+    rc = 0
+    for name in TOOLS:
+        canon = REPO / "config" / "mimo-tools" / name
+        target = Path.home() / ".mimocode" / "tools" / name
+        if not canon.is_file():
+            print(f"{name}: canonical source missing: {canon}")
+            rc = 1
+            continue
+        canon_hash = sha256(canon)
+        if target.is_file() and sha256(target) == canon_hash:
+            print(f"{name}: in-sync {canon_hash[:12]}")
+            continue
+        drift = f"DRIFT canonical={canon_hash[:12]}"
+        if target.is_file():
+            drift += f" installed={sha256(target)[:12]}"
+        else:
+            drift += " installed=MISSING"
+        print(f"{name}: {drift} — installing")
+        target.parent.mkdir(parents=True, exist_ok=True)
+        shutil.copy2(canon, target)
+        if sha256(target) != canon_hash:
+            print(f"{name}: INSTALL FAILED (hash mismatch after copy)")
+            rc = 1
+        else:
+            print(f"{name}: installed {canon_hash[:12]}")
+    return rc
+
+
+if __name__ == "__main__":
+    sys.exit(main())
@@ -0,0 +1,83 @@
+#!/usr/bin/env python3
+"""install_wr_daemon.py — REAL_ASYNC_BRIDGE §6: daemon governance.
+
+Canonical source lives IN THIS REPO at bridge/daemon/. The active install
+is ~/.local/share/mimo-wait-resume/. This script is the ONLY supported
+way to move code between them:
+
+  - copy canonical source → install dir (atomic, per file)
+  - verify sha256 convergence (fail loud on divergence)
+  - optionally restart the daemon via scripts/start_wr_daemon.py
+
+Never hand-edit the installed copy: edit the repo source, install, restart.
+"""
+
+from __future__ import annotations
+
+import argparse
+import hashlib
+import shutil
+import subprocess
+import sys
+from pathlib import Path
+
+REPO = Path(__file__).resolve().parent.parent
+CANONICAL = REPO / "bridge" / "daemon"
+INSTALL = Path.home() / ".local/share/mimo-wait-resume"
+FILES = ("daemon.py", "waitstore.py", "verify.py", "completion_file.py")
+
+
+def sha256(p: Path) -> str:
+    return hashlib.sha256(p.read_bytes()).hexdigest()
+
+
+def main() -> int:
+    ap = argparse.ArgumentParser(description=__doc__)
+    ap.add_argument("--check", action="store_true", help="verify convergence only; no writes")
+    ap.add_argument("--restart", action="store_true", help="restart the daemon after install")
+    args = ap.parse_args()
+
+    if not CANONICAL.is_dir():
+        print(f"canonical source missing: {CANONICAL}")
+        return 1
+    INSTALL.mkdir(parents=True, exist_ok=True)
+
+    diverged = []
+    for name in FILES:
+        src, dst = CANONICAL / name, INSTALL / name
+        if args.check:
+            if not dst.is_file() or sha256(src) != sha256(dst):
+                diverged.append(name)
+            continue
+        if not dst.is_file() or sha256(src) != sha256(dst):
+            tmp = dst.with_suffix(f".tmp.{name}")
+            shutil.copy2(src, tmp)
+            tmp.replace(dst)  # atomic
+            print(f"installed: {name} {sha256(src)[:12]}")
+        else:
+            print(f"unchanged: {name} {sha256(src)[:12]}")
+
+    if args.check:
+        if diverged:
+            print(f"DIVERGED: {', '.join(diverged)} — run install_wr_daemon.py")
+            return 1
+        print("CONVERGED: installed copy == canonical repo source")
+        return 0
+
+    # post-install convergence proof
+    bad = [n for n in FILES if sha256(CANONICAL / n) != sha256(INSTALL / n)]
+    if bad:
+        print(f"INSTALL FAILED convergence: {bad}")
+        return 1
+    print("CONVERGED: installed copy == canonical repo source")
+
+    if args.restart:
+        rc = subprocess.call([sys.executable, str(REPO / "scripts" / "start_wr_daemon.py")])
+        if rc != 0:
+            print("restart failed")
+            return 1
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
@@ -27,6 +27,7 @@ capture time, the caller must decide (abort or accept-and-fingerprint).
 This module fingerprints regardless, so even a dirty worktree can't hide
 in-stage mutations.
 """
+
 from __future__ import annotations

 import hashlib
@@ -39,8 +40,7 @@ class MutationError(RuntimeError):


 def _git(*args: str, cwd: str) -> str:
-    result = subprocess.run(["git", *args], cwd=cwd,
-                            capture_output=True, text=True, timeout=30)
+    result = subprocess.run(["git", *args], cwd=cwd, capture_output=True, text=True, timeout=30)
    return result.stdout.strip()


@@ -56,8 +56,7 @@ def capture_worktree_state(repo_root: str) -> dict:
        state["head"] = _git("rev-parse", "HEAD", cwd=repo_root)
        state["worktree_clean"] = _git("status", "--porcelain", cwd=repo_root) == ""
        # All files (tracked + untracked, excluding .git)
-        files = _git("ls-files", "--cached", "--others", "--exclude-standard",
-                     cwd=repo_root)
+        files = _git("ls-files", "--cached", "--others", "--exclude-standard", cwd=repo_root)
        fingerprints = {}
        for f in files.splitlines():
            fp = os.path.join(repo_root, f)
@@ -73,8 +72,7 @@ def capture_worktree_state(repo_root: str) -> dict:
    return state


-def verify_no_source_mutation(repo_root: str, before: dict,
-                              allowed_paths: list[str]) -> list[str]:
+def verify_no_source_mutation(repo_root: str, before: dict, allowed_paths: list[str]) -> list[str]:
    """D5.1+D5.2: content-hash comparison + explicit allowlist.

    Returns violations (empty = clean).
@@ -104,8 +102,7 @@ def verify_no_source_mutation(repo_root: str, before: dict,
    def _is_allowed(rel: str) -> bool:
        if rel in allowed_exact:
            return True
-        return any(rel.startswith(prefix + os.sep) or rel.startswith(prefix + "/")
-                   for prefix in allowed_prefixes)
+        return any(rel.startswith(prefix + os.sep) or rel.startswith(prefix + "/") for prefix in allowed_prefixes)

    # D5.1: compare fingerprints — detects changes to already-dirty files
    for path in sorted(set(before_fp) | set(after_fp)):
@@ -122,8 +119,7 @@ def verify_no_source_mutation(repo_root: str, before: dict,

    # HEAD must not move (cognitive stages must never commit)
    if before.get("head") and after.get("head") and before["head"] != after["head"]:
-        violations.append(
-            f"UNAUTHORIZED_COMMIT: HEAD moved {before['head'][:8]} → {after['head'][:8]}")
+        violations.append(f"UNAUTHORIZED_COMMIT: HEAD moved {before['head'][:8]} → {after['head'][:8]}")

    return violations

@@ -0,0 +1,102 @@
+#!/usr/bin/env python3
+"""notify_event_bridge.py — T16 §16C: ci.yml's final job posts the CI
+completion WAKE to the local W/R event bridge (daemon /process-event).
+
+The event is a WAKE, never completion truth: the waiting controller
+(ci_event_wait.py) re-reads the canonical Gitea verdict via
+deploy_qualify.py after resuming. Best-effort by design — if the bridge
+is unreachable the waiter's watchdog still recovers truthfully (that is
+the §16D lost-event contract), so a notify failure must NEVER fail CI.
+
+REAL_ASYNC_BRIDGE (R3): the completion event is ALSO written directly as
+a durable completion file (resumes/<operation_id>.done). Delivery no
+longer depends on the daemon being alive: the waiter blocks on inotify
+for the file itself (kernel wake, zero latency, zero CPU), and the
+durable file survives daemon AND waiter restarts (R9). The write is
+best-effort on the same terms as the POST — never fails CI.
+"""
+from __future__ import annotations
+
+import json
+import os
+import sys
+import time
+import urllib.request
+
+BRIDGE_URL = os.environ.get("RUNE_EVENT_BRIDGE_URL",
+                            "http://127.0.0.1:8791")
+
+
+def main() -> int:
+    sha = os.environ.get("CI_SHA", "")
+    result = os.environ.get("CI_RESULT", "")  # needs.<job>.result
+    if not sha:
+        print("notify-event-bridge: no CI_SHA — nothing to wake", flush=True)
+        return 0
+    payload = {
+        "operation_id": f"ci-{sha}",
+        "provider": "gitea",
+        "event_type": "ci_completed",
+        "sha": sha,
+        # exit_code is informational only; the wake is not the verdict
+        "exit_code": 0 if result == "success" else 1,
+        "completed_at": time.time(),
+    }
+    # R8 lost-event failure-recovery: operation-scoped fault injection.
+    # A rule keyed by the EXACT operation_id suppresses BOTH normal
+    # delivery channels (durable completion file AND bridge POST), so the
+    # waiter's watchdog/failure-detector path is the only route to truth.
+    # Disabled by default (no rule file => zero behavior change) and
+    # observable: every decision is recorded in the R8 fault log.
+    try:
+        import r8_fault  # noqa: E402 — sibling module
+        if r8_fault.should_suppress(payload["operation_id"]):
+            r8_fault.record(payload["operation_id"], "suppressed",
+                            {"sha": sha, "exit_code": payload["exit_code"],
+                             "source_terminal_at": payload["completed_at"]})
+            print("notify-event-bridge: R8 fault injection — completion "
+                  f"event intentionally suppressed for "
+                  f"{payload['operation_id']} (no .done file, no POST); "
+                  "watchdog authority recovery is the expected path",
+                  flush=True)
+            return 0
+        r8_fault.record(payload["operation_id"], "delivered_normally",
+                        {"sha": sha, "rule_active": False})
+    except Exception as exc:  # noqa: BLE001 — fault seam must never break delivery
+        print(f"notify-event-bridge: R8 fault-seam probe failed ({exc}); "
+              "continuing with normal delivery", flush=True)
+    # R3 durable delivery: write the completion file FIRST — it works even
+    # if the bridge daemon is down, and the atomic rename means any waiter
+    # (inotify-watching or starting later) sees a complete file or none.
+    try:
+        sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
+        import completion_file as cf  # noqa: E402 — sibling module
+        written = cf.write_completion(payload["operation_id"], payload)
+        print(f"notify-event-bridge: durable completion file {written}",
+              flush=True)
+    except Exception as exc:  # noqa: BLE001 — delivery is best-effort; watchdog recovery covers a lost completion file (§16D contract)
+        print(f"notify-event-bridge: completion-file write failed ({exc}); "
+              f"bridge POST and watchdog recovery still cover delivery",
+              flush=True)
+    for attempt in range(2):
+        try:
+            req = urllib.request.Request(
+                BRIDGE_URL + "/process-event",
+                data=json.dumps(payload).encode(),
+                headers={"Content-Type": "application/json"})
+            with urllib.request.urlopen(req, timeout=5) as r:
+                resp = json.loads(r.read())
+            print(f"notify-event-bridge: woke bridge "
+                  f"(matched={resp.get('matched')})", flush=True)
+            return 0
+        except Exception as exc:  # noqa: BLE001 — wake is best-effort by design; failure is logged and the waiter's watchdog recovery covers a lost wake (§16D contract)
+            if attempt == 0:
+                time.sleep(2)
+            else:
+                print(f"notify-event-bridge: bridge unreachable ({exc}); "
+                      f"watchdog recovery covers a lost wake", flush=True)
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
@@ -0,0 +1,193 @@
+#!/usr/bin/env python3
+"""pl1_records.py — PL1-01/PL1-02: structured Experience records from live
+operator outcomes (handoff 2026-09-11, PL1 first tranche).
+
+Enriches the PL0 process_experience ledger with the fields PL1 needs for
+trustworthy learning, sourced from OBSERVABLE task state only:
+
+  task_family      — code|research|chat|general (existing router vocabulary)
+  route            — LOCAL_FIX | STANDARD_CHANGE | RESEARCHED_CHANGE (route.jsonl ref)
+  result_state     — completed | failed | research_required | awaiting_approval
+  proof_level      — verification evidence class: exit_code|test|mutation|none
+  failure_class    — NOT_COMPLETED_NO_MUTATION_EVIDENCE | verification_failure |
+                     research_dispatch | model_refusal | UNKNOWN
+  retry_count      — observed re-issues of the same objective (0 if unknown)
+
+Completion-truth contract respected: a mutation task with no mutating tool
+evidence records failure_class=NOT_COMPLETED_NO_MUTATION_EVIDENCE and is
+NEVER eligible for promotion. Most records stay evidence-only; a record
+becomes a CANDIDATE lesson only via qualify_experience's evidence signals
+PLUS a concrete structural root cause (not UNKNOWN).
+"""
+
+from __future__ import annotations
+
+import json
+import os
+import sys
+from pathlib import Path
+
+_SCRIPTS = os.path.dirname(os.path.abspath(__file__))
+if _SCRIPTS not in sys.path:
+    sys.path.insert(0, _SCRIPTS)
+
+LEDGER_DEFAULT = os.path.expanduser("~/.rune/state/process_experience.jsonl")
+"""Default path for the PL0 Experience Ledger. Override via
+RUNE_EXPERIENCE_LEDGER env var (consumed by process_experience.py)."""
+TASK_STATE_DEFAULT = os.path.expanduser("~/.rune/hot/task_state.json")
+ROUTE_LOG_DEFAULT = os.path.expanduser("~/.rune/routes/route.jsonl")
+
+FAILURE_CLASSES = {
+    "NOT_COMPLETED_NO_MUTATION_EVIDENCE",
+    "verification_failure",
+    "research_dispatch",
+    "model_refusal",
+    "UNKNOWN",
+}
+
+
+def resolve_route(task_id: str, route_ref: str, route_log: str = ROUTE_LOG_DEFAULT) -> str:
+    """Resolve 'route:<decided_at>' into the actual route name from the
+    append-only route log. Returns '' when unresolvable — never guesses."""
+    if not route_ref.startswith("route:"):
+        return ""
+    ts = route_ref[len("route:") :]
+    p = Path(route_log)
+    if not p.exists():
+        return ""
+    try:
+        for line in p.read_text().splitlines():
+            try:
+                ev = json.loads(line)
+            except json.JSONDecodeError:
+                continue
+            if (
+                str(ev.get("decided_at", "")).startswith(ts[:19])
+                and ev.get("dispatch", {}).get("task_id", task_id) == task_id
+            ):
+                return str(ev.get("route", ""))
+    except OSError:
+        pass
+    return ""
+
+
+def proof_level(task: dict) -> str:
+    """Observable verification evidence class for one finished task."""
+    outcome_ref = task.get("outcome_ref") or ""
+    result = task.get("result") or ""
+    if outcome_ref.startswith("response:") and "wrote" in result.lower():
+        return "mutation"
+    # task.outcome journal events carry verification_passed — surfaced in
+    # task.result excerpt matching is too weak; the outcome event is the
+    # authority but reading the whole journal per record is expensive, so
+    # classify from what the checkpoint honestly carries:
+    return "none"
+
+
+def failure_class(task: dict) -> str:
+    """Evidence-grounded failure classification (PL1-02 vocabulary)."""
+    status = task.get("status", "")
+    result = task.get("result") or ""
+    if status == "research_required" or task.get("outcome_ref") == "research_dispatch":
+        return "research_dispatch"
+    if result.startswith("NOT_COMPLETED_NO_MUTATION_EVIDENCE"):
+        return "NOT_COMPLETED_NO_MUTATION_EVIDENCE"
+    if status == "failed":
+        if "verification" in result.lower():
+            return "verification_failure"
+        if "capability" in result.lower() or "refus" in result.lower():
+            return "model_refusal"
+        return "UNKNOWN"
+    return "UNKNOWN"
+
+
+def enrich_record(task: dict, route_log: str = ROUTE_LOG_DEFAULT) -> dict:
+    """Build the PL1 evidence record for one task from canonical state."""
+    route_ref = task.get("route_ref") or ""
+    route = resolve_route(task.get("task_id", ""), route_ref, route_log)
+    status = task.get("status", "")
+    fc = failure_class(task)
+    return {
+        "task_id": task.get("task_id", ""),
+        "task_family": task.get("task_family", ""),  # trace v2 field; '' if absent
+        "route": route,
+        "objective": (task.get("title") or "")[:200],
+        "result_state": status,
+        "proof_level": proof_level(task),
+        "failure_class": fc if status not in ("completed",) else "",
+        "model": task.get("assigned_model") or "",
+        "outcome_ref": task.get("outcome_ref") or "",
+        "human_correction": False,  # set when the user corrects a result
+        "retry_count": 0,  # amended tasks increment; single-pass default
+        "candidate_status": "CANDIDATE" if fc not in ("UNKNOWN", "") else "EVIDENCE_ONLY",
+    }
+
+
+def build_eligible_records(
+    task_state_path: str = TASK_STATE_DEFAULT, route_log: str = ROUTE_LOG_DEFAULT, limit: int = 200
+) -> list[dict]:
+    """PL1-01: compact structured evidence for recent finished tasks."""
+    p = Path(task_state_path)
+    if not p.exists():
+        return []
+    checkpoint = json.loads(p.read_text())
+    tasks = checkpoint.get("tasks", [])
+    seen: set[str] = set()
+    out = []
+    for t in reversed(tasks[-limit:]):
+        tid = t.get("task_id", "")
+        if not tid or tid in seen:
+            continue
+        seen.add(tid)
+        if t.get("status") not in ("completed", "failed", "research_required"):
+            continue
+        out.append(enrich_record(t, route_log))
+    return out
+
+
+def metrics(records: list[dict]) -> dict:
+    """PL1-05: correctness-first metrics from enriched records.
+
+    Evidence-backed semantics (handoff 2026-09-11 §I):
+    - first_pass_accepted: NOT YET MEASURABLE — requires attempt-count data
+      that the current task state does not record (single-pass vs amended).
+    - false_completion_rate: NOT_YET_MEASURABLE — requires post-completion
+      evidence that a 'completed' task was actually not complete. The
+      structural gate (NOT_COMPLETED_NO_MUTATION_EVIDENCE) prevents the
+      KNOWN pathway going forward, but historical records predate it.
+    - known_false_completions: count of records where post-hoc evidence
+      proved non-completion (currently 0 — the structural gate prevents
+      new ones, and historical ones have not been audited).
+    - known_honest_failures: records with evidence-grounded failure classes.
+    - unknown_validity: records whose completion status cannot be verified
+      from available evidence.
+    """
+    total = len(records)
+    completed = [r for r in records if r["result_state"] == "completed"]
+    failed = [r for r in records if r["result_state"] == "failed"]
+    mutation_failures = [r for r in failed if r.get("failure_class") == "NOT_COMPLETED_NO_MUTATION_EVIDENCE"]
+    dispatched = [r for r in records if r.get("failure_class") == "research_dispatch"]
+    candidates = [r for r in records if r.get("candidate_status") == "CANDIDATE"]
+    known_honest_failures = len([r for r in failed if r.get("failure_class") not in ("", "UNKNOWN")])
+    return {
+        "total_finished": total,
+        "completed": len(completed),
+        "failed": len(failed),
+        "first_pass_accepted": "NOT_YET_MEASURABLE",
+        "false_completion_rate": "NOT_YET_MEASURABLE",
+        "known_false_completions": 0,
+        "known_honest_failures": known_honest_failures,
+        "unknown_validity": len(completed),
+        "mutation_failures": len(mutation_failures),
+        "research_dispatches": len(dispatched),
+        "verification_failures": sum(1 for r in failed if r.get("failure_class") == "verification_failure"),
+        "recurring_failure_families": sorted(
+            {r["failure_class"] for r in failed + dispatched if r.get("failure_class") not in ("", "UNKNOWN")}
+        ),
+        "candidates": len(candidates),
+        "evidence_only": total - len(candidates),
+    }
+
+
+if __name__ == "__main__":
+    print(json.dumps(metrics(build_eligible_records()), indent=2))
@@ -13,6 +13,7 @@ publication fails visibly, must not silently falsify the ledger). It does
 NOT alter research cognition or stage outcomes — callers invoke it after
 gate acceptance, and its failure does not invalidate the accepted artifact.
 """
+
 from __future__ import annotations

 import json
@@ -20,17 +21,24 @@ import os
 import urllib.request

 from source_ledger import build_records, render_markdown
+import sys as _sys
+_SCRIPT_DIR = str(__import__('pathlib').Path(__file__).resolve().parent)
+if _SCRIPT_DIR not in _sys.path:
+    _sys.path.insert(0, _SCRIPT_DIR)
+import gitea_auth

 MARKER_PREFIX = "<!-- source-ledger"


 def _api(method: str, path: str, payload: dict | None = None) -> tuple[int, dict | str]:
-    token = os.environ.get("GITEA_TOKEN", "")
+    token = gitea_auth.get_token() if gitea_auth.has_token() else ""
    server = os.environ.get("GITEA_SERVER", "https://gitea.thingscouldgetdicey.online")
    req = urllib.request.Request(
-        server + "/api/v1" + path, method=method,
+        server + "/api/v1" + path,
+        method=method,
        data=json.dumps(payload).encode() if payload else None,
-        headers={"Authorization": f"token {token}", "Content-Type": "application/json"})
+        headers={"Authorization": f"token {token}", "Content-Type": "application/json"},
+    )
    try:
        with urllib.request.urlopen(req, timeout=60) as r:
            body = r.read().decode()
@@ -43,8 +51,9 @@ def _marker(artifact_hash: str) -> str:
    return f"{MARKER_PREFIX} hash={artifact_hash} -->"


-def publish(issue: int, artifact_dir: str, artifact_hash: str,
-            runner: dict | None = None, model: str | None = None) -> dict:
+def publish(
+    issue: int, artifact_dir: str, artifact_hash: str, runner: dict | None = None, model: str | None = None
+) -> dict:
    """Render + post/update the ledger comment. Returns publication result."""
    path = os.path.join(artifact_dir, "outward.json")
    with open(path) as f:
@@ -58,9 +67,13 @@ def publish(issue: int, artifact_dir: str, artifact_hash: str,
    legacy = not any(
        c.get("contradicts_source_ids")
        for c in outward.get("contradictory_evidence", outward.get("contradictions", []))
-        if isinstance(c, dict))
-    body = _marker(artifact_hash) + "\n" + render_markdown(
-        records, outward, runner, model, artifact_hash, legacy_claims=legacy)
+        if isinstance(c, dict)
+    )
+    body = (
+        _marker(artifact_hash)
+        + "\n"
+        + render_markdown(records, outward, runner, model, artifact_hash, legacy_claims=legacy)
+    )

    # Find existing ledger comment for this artifact hash (idempotency)
    status, comments = _api("GET", f"/repos/conductor/rune/issues/{issue}/comments?limit=50")
@@ -73,14 +86,12 @@ def publish(issue: int, artifact_dir: str, artifact_hash: str,
            break

    if existing:
-        status, _ = _api("PATCH", f"/repos/conductor/rune/issues/comments/{existing['id']}",
-                         {"body": body})
+        status, _ = _api("PATCH", f"/repos/conductor/rune/issues/comments/{existing['id']}", {"body": body})
        if status not in (200, 201):
            raise RuntimeError(f"SourceLedger publish: update failed (HTTP {status})")
        return {"action": "updated", "comment_id": existing["id"], "artifact_hash": artifact_hash}

-    status, created = _api("POST", f"/repos/conductor/rune/issues/{issue}/comments",
-                           {"body": body})
+    status, created = _api("POST", f"/repos/conductor/rune/issues/{issue}/comments", {"body": body})
    if status not in (200, 201):
        raise RuntimeError(f"SourceLedger publish: create failed (HTTP {status}): {created}")
    return {"action": "created", "comment_id": created.get("id"), "artifact_hash": artifact_hash}
@@ -88,6 +99,7 @@ def publish(issue: int, artifact_dir: str, artifact_hash: str,

 def main_with(argv: list[str]) -> int:
    import argparse
+
    parser = argparse.ArgumentParser(description="Publish the OUTWARD Source Ledger")
    parser.add_argument("--issue", type=int, required=True)
    parser.add_argument("--artifact-dir", required=True)
@@ -100,4 +112,5 @@ def main_with(argv: list[str]) -> int:

 if __name__ == "__main__":
    import sys
+
    raise SystemExit(main_with(sys.argv))
@@ -0,0 +1,83 @@
+#!/usr/bin/env python3
+"""r8_fault.py — REAL_ASYNC_BRIDGE R8: operation-scoped lost-event fault
+injection (evidence-only harness).
+
+DESIGN CONTRACT (R8 §3):
+  disabled by default          no rule file -> zero behavior change
+  operation-scoped             rule keyed by EXACT operation_id
+  reversible                   delete the rule file / rule entry
+  observable                   every decision appended to a JSONL log
+  cannot affect other waits    rule file checked by operation_id only
+
+The ONLY consumer is scripts/notify_event_bridge.py: before normal
+delivery it consults should_suppress(operation_id). Suppression means:
+  - NO durable completion file is written
+  - NO bridge POST (no /process-event wake)
+  - the suppression decision is recorded as R8 fault evidence
+Nothing else in the stack reads the rule. The daemon, waitstore,
+authority verification, and completion-file machinery are untouched.
+"""
+
+from __future__ import annotations
+
+import json
+import os
+import sys
+import time
+from pathlib import Path
+
+RUNE_WR_ROOT = os.environ.get("RUNE_WR_ROOT") or os.environ.get("WR_ROOT") or "/mnt/workdrive/wr-bridge"
+RULES_PATH = Path(RUNE_WR_ROOT) / "r8_suppress_rules.json"
+FAULT_LOG = Path(RUNE_WR_ROOT) / "r8_fault_log.jsonl"
+
+REASON = "event intentionally suppressed for R8 lost-event failure-recovery"
+
+
+def _rules() -> dict:
+    """Read the rule file. Absent/unreadable => no suppression (fail-open
+    to NORMAL delivery; the fault must never take the healthy path down)."""
+    try:
+        return json.loads(RULES_PATH.read_text())
+    except Exception:  # noqa: BLE001 — absent rule file is the default state
+        return {}
+
+
+def lookup(operation_id: str) -> dict | None:
+    """Return the active suppression rule for EXACTLY this operation_id,
+    or None. Expired rules are ignored (not deleted — deletion is the
+    operator's explicit disable action)."""
+    spec = _rules()
+    if not spec.get("enabled"):
+        return None
+    rule = spec.get("operations", {}).get(operation_id)
+    if not rule:
+        return None
+    expires = rule.get("expires_at")
+    if expires:
+        try:
+            if time.time() > time.mktime(time.strptime(expires, "%Y-%m-%dT%H:%M:%SZ")) - time.timezone:
+                return None
+        except ValueError:
+            return None
+    return rule
+
+
+def should_suppress(operation_id: str) -> bool:
+    return lookup(operation_id) is not None
+
+
+def record(operation_id: str, stage: str, extra: dict | None = None) -> None:
+    """Append a durable fault-evidence line (best-effort, never raises)."""
+    try:
+        entry = {
+            "at": time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime()),
+            "operation_id": operation_id,
+            "reason": REASON,
+            "stage": stage,
+            **(extra or {}),
+        }
+        FAULT_LOG.parent.mkdir(parents=True, exist_ok=True)
+        with FAULT_LOG.open("a") as f:
+            f.write(json.dumps(entry, sort_keys=True) + "\n")
+    except OSError as exc:  # noqa: BLE001 — evidence logging must never break CI
+        print(f"r8_fault: evidence log write failed: {exc}", file=sys.stderr)
@@ -0,0 +1,155 @@
+#!/usr/bin/env python3
+"""release_history_cli.py — thin workflow entry points for release history.
+
+SCAFFOLD-TRANCHE-009B §4: the workflows invoke THIS tested module; no
+note-building logic lives in YAML. Two moments:
+
+    init   RELEASE INITIALIZATION — auto-version creates the evidence-
+           derived Gitea release when the version/tag appears.
+    refresh  RELEASE CONVERGENCE REFRESH — deploy promotes a later SHA of
+           the same version; the SAME release is updated with the final
+           deployed identity + newly-governed facts. Idempotent.
+
+Fail-closed boundary (§5/§16): `init` exits non-zero on publication
+failure — Rune cannot silently create another empty historical release.
+`refresh` ALSO exits non-zero (visible failure + retry converges), but the
+workflow marks it non-blocking for deployment validity: valid software is
+never rolled back over documentation. Failure state is visible either way.
+"""
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import sys
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent / "src"))
+from rune import release_notes as rn  # noqa: E402
+
+
+def _token() -> str:
+    tok = os.environ.get("GITEA_TOKEN", "")
+    if tok:
+        return tok
+    # local/dev fallback: git credential store (runner uses GITEA_TOKEN)
+    cred = Path.home() / ".git-credentials"
+    if cred.exists():
+        import re
+        m = re.search(r"https://[^:]+:([^@]+)@", cred.read_text())
+        if m:
+            return m.group(1)
+    return ""
+
+
+def main(argv=None) -> int:
+    ap = argparse.ArgumentParser(description=__doc__)
+    sub = ap.add_subparsers(dest="cmd", required=True)
+
+    p_init = sub.add_parser("init", help="release initialization")
+    p_init.add_argument("--version", required=True)
+    p_init.add_argument("--server", required=True)
+    p_init.add_argument("--repo", required=True)
+    p_init.add_argument("--store", default="store/scaffold/current.json")
+    p_init.add_argument("--release-sha", default="")
+    p_init.add_argument("--deployed-sha", default="")
+
+    p_ref = sub.add_parser("refresh", help="convergence refresh")
+    p_ref.add_argument("--version", required=True)
+    p_ref.add_argument("--server", required=True)
+    p_ref.add_argument("--repo", required=True)
+    p_ref.add_argument("--store", default="store/scaffold/current.json")
+    p_ref.add_argument("--release-sha", default="")
+    p_ref.add_argument("--deployed-sha", required=True,
+                       help="converged deployed commit (from the deployment "
+                            "authority — required, never assumed)")
+
+    args = ap.parse_args(argv)
+    token = _token()
+    if not token:
+        print(json.dumps({"failed": True,
+                          "error": "no Gitea token available"}))
+        return 1
+
+    if args.cmd == "init":
+        note = rn.auto_initial_note(
+            version=args.version,
+            summary=f"Rune v{args.version} — evidence-derived release "
+                    f"history (automatic initialization)",
+            scaffold_store=args.store,
+            release_sha=args.release_sha,
+            deployed_sha=args.deployed_sha or args.release_sha,
+        )
+        result = rn.publish_release(args.server, args.repo, token, note)
+        print(json.dumps(result))
+        # fail-closed: initialization failure must be visible so Rune cannot
+        # silently create another empty historical release (§5/§16)
+        return 0 if not result.get("failed") else 1
+
+    if args.cmd == "refresh":
+        # §12C RACE-SAFE REVALIDATION: between promotion and refresh, a
+        # NEWER deploy may have superseded this candidate. The deployment
+        # authority is re-read NOW: if current's build identity differs
+        # from the candidate, this job is stale → refuse (preserve T11
+        # invariant: stale jobs never overwrite release history). When
+        # current AGREES, the candidate IS the converged truth.
+        current_link = os.environ.get(
+            "RUNE_DEPLOY_CURRENT", "/mnt/workdrive/rune-code/current")
+        try:
+            deployed_now = rn.converged_deployed_sha(
+                str(Path(current_link).parent))
+        except (OSError, ValueError):
+            deployed_now = ""
+        if args.deployed_sha and deployed_now and \
+                not deployed_now.startswith(args.deployed_sha[:12]):
+            print(json.dumps({
+                "failed": True, "action": "refresh-superseded",
+                "error": f"deployment authority shows {deployed_now[:12]}, "
+                         f"candidate {args.deployed_sha[:12]} was superseded "
+                         f"— stale refresh refused (history preserved)",
+                "authority": deployed_now[:12],
+            }))
+            return 1  # visible; retry by the WINNING deploy converges
+
+        existing = rn.fetch_release(args.server, args.repo, token,
+                                    f"v{args.version}")
+        if existing is None:
+            # no release yet for this version — init instead (convergence
+            # may legitimately arrive before/without an init if the
+            # auto-version run failed); idempotent recovery path
+            note = rn.build_from_deployment(
+                version=args.version, deployed_sha=args.deployed_sha,
+                scaffold_store=args.store, release_sha=args.release_sha)
+        else:
+            note = rn.convergence_refresh(
+                rn.ReleaseNote.from_dict({
+                    "version": args.version,
+                    "summary": _summary_from_body(existing.get("body", "")),
+                    "provenance": [],
+                }),
+                deployed_sha=args.deployed_sha, scaffold_store=args.store,
+                extra_provenance="automatic convergence refresh (deploy "
+                                 "workflow)")
+            # preserve structural fields the stored body carries
+            note.release_sha = _sha_from_body(existing.get("body", ""),
+                                              "tag") or args.release_sha
+        result = rn.publish_release(args.server, args.repo, token, note)
+        print(json.dumps(result))
+        # visible failure; retry converges (§10). Deployment validity is
+        # NOT rolled back over documentation — the workflow marks this
+        # step appropriately.
+        return 0 if not result.get("failed") else 1
+    return 2
+
+
+def _summary_from_body(body: str) -> str:
+    """Delegate to the single authority in rune.release_notes (§14B)."""
+    return rn.summary_from_body(body)
+
+
+def _sha_from_body(body: str, which: str) -> str:
+    return rn.sha_from_body(body, which)
+
+
+if __name__ == "__main__":
+    sys.exit(main())
@@ -9,6 +9,7 @@ DEFINE does NOT: inspect code, search externally, or produce a design.
 DEFINE does: classify the question, identify sub-questions, define the evidence
 contract, and produce the required-stage plan.
 """
+
 from __future__ import annotations

 import json
@@ -17,15 +18,36 @@ from typing import Any

 # Keywords that suggest external research is needed
 EXTERNAL_INDICATORS = [
-    "best practices", "state of the art", "alternative", "compare",
-    "benchmark", "library", "framework", "prior art", "existing solutions",
-    "how do other", "state of the art", "literature", "approach",
+    "best practices",
+    "state of the art",
+    "alternative",
+    "compare",
+    "benchmark",
+    "library",
+    "framework",
+    "prior art",
+    "existing solutions",
+    "how do other",
+    "state of the art",
+    "literature",
+    "approach",
 ]

 # Keywords that suggest repo-specific work
 REPO_INDICATORS = [
-    "fix", "bug", "refactor", "add", "implement", "remove", "optimize",
-    "update", "migrate", "clean up", "wire", "integrate", "test",
+    "fix",
+    "bug",
+    "refactor",
+    "add",
+    "implement",
+    "remove",
+    "optimize",
+    "update",
+    "migrate",
+    "clean up",
+    "wire",
+    "integrate",
+    "test",
 ]


@@ -88,8 +110,7 @@ def compile_define_contract(issue: dict) -> dict:
    required_stages.extend(["SYNTHESIS", "QUALIFY"])

    # Skipped stages with reasons (canonical stage names only)
-    all_stages = {"ISSUE", "INWARD", "OUTWARD", "APPLICABILITY",
-                  "ADVERSARIAL", "INNOVATION", "SYNTHESIS", "QUALIFY"}
+    all_stages = {"ISSUE", "INWARD", "OUTWARD", "APPLICABILITY", "ADVERSARIAL", "INNOVATION", "SYNTHESIS", "QUALIFY"}
    skipped = {}
    for s in sorted(all_stages - set(required_stages)):
        if s == "OUTWARD":
@@ -144,6 +165,7 @@ def is_ambiguous(issue: dict) -> bool:

 def main_with(argv: list[str]) -> int:
    import argparse
+
    parser = argparse.ArgumentParser(description="Deterministic DEFINE stage")
    parser.add_argument("--issue-file", required=True, help="Path to issue JSON")
    parser.add_argument("--out", required=True, help="Output path for define.json")
@@ -163,12 +185,15 @@ def main_with(argv: list[str]) -> int:
    with open(args.out, "w") as f:
        json.dump(result, f, indent=2, default=str)

-    print(f"DEFINE contract: type={contract['question_type']}, "
-          f"stages={contract['required_stages']}, "
-          f"mimo_needed={ambiguous}")
+    print(
+        f"DEFINE contract: type={contract['question_type']}, "
+        f"stages={contract['required_stages']}, "
+        f"mimo_needed={ambiguous}"
+    )
    return 0


 if __name__ == "__main__":
    import sys
+
    sys.exit(main_with(sys.argv))
@@ -8,6 +8,7 @@ Event types emitted:
  StageStarted, PromptCompiled, QueryIssued, ToolStarted, ToolCompleted,
  SourceDiscovered, SourceRead, EvidenceAccepted, ToolFailed, StageCompleted, StageFailed
 """
+
 from __future__ import annotations

 import json
@@ -21,8 +22,7 @@ from typing import Callable
 class ResearchEventRelay:
    """Reads mimo JSON events in real-time and updates canonical state."""

-    def __init__(self, tree, events_path: str, mimo_cmd: list[str], err_path: str,
-                 attempt=None):
+    def __init__(self, tree, events_path: str, mimo_cmd: list[str], err_path: str, attempt=None):
        self.tree = tree
        self.events_path = events_path
        self.mimo_cmd = mimo_cmd
@@ -31,8 +31,12 @@ class ResearchEventRelay:
        # marks tool windows; it never changes classification or control flow.
        self.attempt = attempt
        self.event_counts = {
-            "queries": 0, "sources_discovered": 0, "sources_read": 0,
-            "evidence_accepted": 0, "tool_failures": 0, "tool_calls": 0,
+            "queries": 0,
+            "sources_discovered": 0,
+            "sources_read": 0,
+            "evidence_accepted": 0,
+            "tool_failures": 0,
+            "tool_calls": 0,
        }
        # Real tool receipts (RP3 evidence chain: QUERY → SOURCE → READ → CLAIM)
        self.receipts = []
@@ -62,31 +66,45 @@ class ResearchEventRelay:
                    url = tool_input["url"]
                    if "duckduckgo" in url or "search" in url:
                        self.event_counts["queries"] += 1
-                        self.receipts.append({
-                            "kind": "QUERY", "tool": tool, "url": url,
-                            "ts": time.time(),
-                        })
+                        self.receipts.append(
+                            {
+                                "kind": "QUERY",
+                                "tool": tool,
+                                "url": url,
+                                "ts": time.time(),
+                            }
+                        )
                        return "QueryIssued"
                    else:
                        self.event_counts["sources_read"] += 1
-                        self.receipts.append({
-                            "kind": "SOURCE_READ", "tool": tool, "url": url,
-                            "ts": time.time(),
-                        })
+                        self.receipts.append(
+                            {
+                                "kind": "SOURCE_READ",
+                                "tool": tool,
+                                "url": url,
+                                "ts": time.time(),
+                            }
+                        )
                        return "SourceRead"
-                self.receipts.append({
-                    "kind": "TOOL", "tool": tool,
-                    "input_keys": sorted(tool_input.keys()) if isinstance(tool_input, dict) else [],
-                    "ts": time.time(),
-                })
+                self.receipts.append(
+                    {
+                        "kind": "TOOL",
+                        "tool": tool,
+                        "input_keys": sorted(tool_input.keys()) if isinstance(tool_input, dict) else [],
+                        "ts": time.time(),
+                    }
+                )
                return "ToolCompleted"
            elif status in ("error", "failed"):
                self.event_counts["tool_failures"] += 1
-                self.receipts.append({
-                    "kind": "TOOL_FAILED", "tool": tool,
-                    "input_keys": sorted(tool_input.keys()) if isinstance(tool_input, dict) else [],
-                    "ts": time.time(),
-                })
+                self.receipts.append(
+                    {
+                        "kind": "TOOL_FAILED",
+                        "tool": tool,
+                        "input_keys": sorted(tool_input.keys()) if isinstance(tool_input, dict) else [],
+                        "ts": time.time(),
+                    }
+                )
                return "ToolFailed"
        return None

@@ -163,11 +181,22 @@ def run_stage_with_relay(tree, out_dir: str) -> dict:
    err_path = os.path.join(out_dir, f"stage_{stage}.mimo.err.log")

    mimo_bin = os.environ.get("MIMO_BIN", "/var/lib/gitea-desktop-runner/bin/mimo")
-    mimo_model = os.environ.get("MIMO_MODEL", "Verboo Code/glm-5.3-flash")
+    from rune.provider_context import require_model
+    mimo_model = require_model()  # T16 §16F: provider layer owns selection
    mimo_agent = os.environ.get("MIMO_AGENT", "build")

-    cmd = [mimo_bin, "run", "--agent", mimo_agent, "-m", mimo_model,
-           "--dangerously-skip-permissions", "--format", "json", prompt]
+    cmd = [
+        mimo_bin,
+        "run",
+        "--agent",
+        mimo_agent,
+        "-m",
+        mimo_model,
+        "--dangerously-skip-permissions",
+        "--format",
+        "json",
+        prompt,
+    ]

    relay = ResearchEventRelay(tree, events_path, cmd, err_path)
    result = relay.run()
@@ -10,6 +10,7 @@ Builds a bounded candidate context for MiMo INWARD inspection:
 This is DETERMINISTIC — no MiMo needed. MiMo gets the candidate context
 and only performs semantic inspection if the context requires it.
 """
+
 from __future__ import annotations

 import json
@@ -37,11 +38,17 @@ def build_context(issue: dict, define_contract: dict) -> dict:
    # 3. Find matching test files
    matched_tests = _find_tests(keywords)

+    # 4. T21.5 §10: Memory internal-experience port — ONE port of call,
+    #    bounded, zero-result = normal (fast path: [] → next step directly).
+    from research_memory_port import memory_port_of_call
+    memory_candidates = memory_port_of_call(issue, keywords)
+
    return {
        "keywords": keywords,
        "umbrella_matches": matched_modules,
        "candidate_files": matched_files[:15],  # bounded
        "candidate_tests": matched_tests[:10],  # bounded
+        "memory_candidates": memory_candidates,  # evidence, never authority
        "define_contract": define_contract,
        "mimo_needed": _needs_mimo(matched_files, matched_tests),
    }
@@ -49,16 +56,44 @@ def build_context(issue: dict, define_contract: dict) -> dict:

 def _extract_keywords(text: str) -> list[str]:
    """Extract meaningful keywords from the issue text."""
-    stop = {"the", "a", "an", "in", "for", "to", "of", "and", "or", "is", "are",
-            "when", "how", "what", "this", "that", "with", "on", "at", "from"}
+    stop = {
+        "the",
+        "a",
+        "an",
+        "in",
+        "for",
+        "to",
+        "of",
+        "and",
+        "or",
+        "is",
+        "are",
+        "when",
+        "how",
+        "what",
+        "this",
+        "that",
+        "with",
+        "on",
+        "at",
+        "from",
+    }
    words = re.findall(r"[a-z_]{3,}", text.lower())
    return [w for w in words if w not in stop][:20]


 def _load_manifest() -> dict:
+    """Structural facts from SPINE (the single structural authority).
+
+    Phase D: reads the Spine-generated compatibility view (same schema as
+    the old manual manifest). The view is OUTPUT — the authority is
+    store/spine/current.json; direct edits to the view are flagged by
+    `spine drift`. If a manual manifest is present (Spine not yet seeded
+    in this checkout), it degrades gracefully to that legacy truth."""
    if os.path.exists(UMBRELLA_MANIFEST):
        with open(UMBRELLA_MANIFEST) as f:
-            return json.load(f)
+            data = json.load(f)
+        return data  # generated view or manual manifest — both work
    return {"umbrellas": {}}


@@ -0,0 +1,139 @@
+"""research_memory_port.py — T21.5 §10-12: the future-facing Memory port
+inside the EXISTING INWARD research stage.
+
+BOUNDARIES (G9/G10/G11):
+  - This is a PORT, not a memory system. No storage is created here; the
+    implementation delegates to a provider callable that future Memory
+    will supply. Tests inject doubles; INWARD never imports Memory.
+  - Zero result is a NORMAL SUCCESSFUL result: the caller continues the
+    normal INWARD flow immediately (no extra model call, no ceremony).
+  - Returned material is EVIDENCE/CANDIDATES only — never an instruction
+    authority. APPLICABILITY (existing stage) governs whether a candidate
+    applies; SYNTHESIS receives only distilled, applicable consequences.
+  - Teachable-moment seam: durable construction evidence can mark an
+    episode as a CANDIDATE for future learning ("may be worth learning",
+    never "always apply"). Substrate/Memory (future) harvest candidates;
+    today the seam only records them via the same provider contract.
+"""
+from __future__ import annotations
+
+import os
+from typing import Any, Callable
+
+# The provider contract: ONE callable, independently replaceable/mockable.
+# Signature: query(query_text, keywords, *, limit) -> list[MemoryCandidate]
+# MemoryCandidate (dict): {source, kind, summary, relevance, episode_ref}
+#   kind ∈ {construction_episode, teachable_moment_candidate,
+#           prior_failure, prior_success}
+# Future Memory supplies the real implementation; nothing else changes.
+
+MemoryProvider = Callable[..., list[dict]]
+
+_DEFAULT_PROVIDER: MemoryProvider | None = None
+
+
+def set_memory_provider(provider: MemoryProvider | None) -> None:
+    """Bind the Memory-side implementation (production or test double).
+    None disables the port — INWARD continues without it (fast path)."""
+    global _DEFAULT_PROVIDER
+    _DEFAULT_PROVIDER = provider
+
+
+def _provider() -> MemoryProvider | None:
+    global _DEFAULT_PROVIDER
+    if _DEFAULT_PROVIDER is not None:
+        return _DEFAULT_PROVIDER
+    env = os.environ.get("RUNE_MEMORY_PORT_PROVIDER")
+    if env:
+        import importlib.util
+        spec = importlib.util.spec_from_file_location("mprov", env)
+        if spec is not None and spec.loader is not None:
+            mod = importlib.util.module_from_spec(spec)
+            spec.loader.exec_module(mod)
+            provider = getattr(mod, "query", None)
+            if provider is not None:
+                _DEFAULT_PROVIDER = provider
+    return _DEFAULT_PROVIDER
+
+
+MAX_CANDIDATES = 5
+MIN_RELEVANCE = 0.3
+
+
+def memory_port_of_call(issue: dict, keywords: list[str],
+                        *, provider: MemoryProvider | None = None) -> list[dict]:
+    """The ONE internal port of call inside INWARD. Bounded retrieval of
+    internal evidence (construction episodes, teachable-moment
+    candidates, prior failures/successes).
+
+    ZERO-RESULT FAST PATH (§11): empty/thin results return [] and the
+    caller continues the next normal INWARD step directly. No extra
+    model call, no fallback reasoning, no synthetic summary, no delay.
+    """
+    prov = provider or _provider()
+    if prov is None:
+        return []  # port unbound: normal INWARD continues, zero overhead
+    try:
+        query_text = " ".join([
+            str(issue.get("title", "")),
+            str(issue.get("body", ""))[:400],
+        ]).strip()
+        raw = prov(query_text, keywords, limit=MAX_CANDIDATES) or []
+    except Exception:  # noqa: BLE001 — port failure must never break INWARD
+        return []
+    candidates = [c for c in raw
+                  if float(c.get("relevance", 0.0)) >= MIN_RELEVANCE]
+    return candidates[:MAX_CANDIDATES]
+
+
+def applicability_filter(candidates: list[dict],
+                         context_keywords: list[str]) -> tuple[list[dict],
+                                                              list[dict]]:
+    """Existing APPLICABILITY governs transfer: candidates relevant to
+    the CURRENT context pass; irrelevant ones are rejected WITHOUT
+    contaminating downstream SYNTHESIS. Returns (applicable, rejected)."""
+    applicable, rejected = [], []
+    for c in candidates:
+        text = (str(c.get("summary", "")) + " " +
+                " ".join(str(k) for k in c.get("keywords", []))).lower()
+        overlap = sum(1 for k in context_keywords if k and k.lower() in text)
+        if overlap > 0 or c.get("explicitly_applicable") is True:
+            applicable.append(c)
+        else:
+            rejected.append(c)
+    return applicable, rejected
+
+
+# ── §12: teachable-moment future seam ──────────────────────────────────────
+
+CANDIDATE_TRIGGERS = (
+    "completed_release", "capability_completed", "failure_recovered",
+    "rollback_executed", "user_correction", "prediction_observation_mismatch",
+    "procedure_improved",
+)
+
+
+def mark_teachable_moment_candidate(
+        episode_ref: str, *, trigger: str, summary: str,
+        emit: Callable[[dict], Any] | None = None) -> dict | None:
+    """Record durable evidence that an episode MAY be worth future
+    learning — a candidate, never a rule. Uses the existing evidence/
+    event mechanism via `emit` (the caller's channel: journal, event
+    bridge, ledger history). No new canonical store is created here;
+    future Substrate/Memory harvests candidates through the port."""
+    if trigger not in CANDIDATE_TRIGGERS:
+        raise ValueError(
+            f"unknown trigger {trigger!r} — must be one of "
+            f"{CANDIDATE_TRIGGERS}")
+    record = {
+        "candidate": "teachable_moment",
+        "episode_ref": episode_ref,
+        "trigger": trigger,
+        "summary": summary[:200],
+        "status": "CANDIDATE",   # never authority, never auto-applied
+        "note": "may be worth future learning; APPLICABILITY governs "
+                "any future transfer",
+    }
+    if emit is not None:
+        emit(record)
+    return record
@@ -12,6 +12,7 @@ and compiles a task-specific prompt that answers:
  What am I forbidden from doing yet?
  What output must I return?
 """
+
 from __future__ import annotations

 import json
@@ -73,6 +74,7 @@ def compile_research_prompt(

 def main_with(argv: list[str]) -> int:
    import argparse
+
    parser = argparse.ArgumentParser(description="Compile a stage-scoped research prompt")
    parser.add_argument("--issue", type=int, required=True)
    parser.add_argument("--round", type=int, default=1)
@@ -92,4 +94,5 @@ def main_with(argv: list[str]) -> int:

 if __name__ == "__main__":
    import sys
+
    sys.exit(main_with(sys.argv))
@@ -35,6 +35,7 @@ import time
 from pathlib import Path

 sys.path.insert(0, str(Path(__file__).parent))
+import gitea_auth  # noqa: E402 — after scripts-dir path insert
 # Dependency direction (handoff 2026-09-10 §3): the research executor MAY
 # import Rune Core; Core never imports the executor. Route authority and
 # evidence probe live in src/rune (canonical, Option C unified authority).
@@ -792,7 +793,8 @@ class StageExecutor:
        prompt = chr(10).join(pp)

        mimo_bin = os.environ.get("MIMO_BIN", "/var/lib/gitea-desktop-runner/bin/mimo")
-        mimo_model = os.environ.get("MIMO_MODEL", "Verboo Code/glm-5.3-flash")
+        from rune.provider_context import require_model
+        mimo_model = require_model()  # T16 §16F: provider layer owns selection
        mimo_agent = os.environ.get("MIMO_AGENT", "build")
        events_path = os.path.join(self.artifact_dir, "stage_OUTWARD.events.jsonl")
        err_path = os.path.join(self.artifact_dir, "stage_OUTWARD.mimo.err.log")
@@ -932,7 +934,8 @@ class StageExecutor:
        prompt = chr(10).join(pp)

        mimo_bin = os.environ.get("MIMO_BIN", "/var/lib/gitea-desktop-runner/bin/mimo")
-        mimo_model = os.environ.get("MIMO_MODEL", "Verboo Code/glm-5.3-flash")
+        from rune.provider_context import require_model
+        mimo_model = require_model()  # T16 §16F: provider layer owns selection
        mimo_agent = os.environ.get("MIMO_AGENT", "build")
        events_path = os.path.join(self.artifact_dir, "stage_APPLICABILITY.events.jsonl")
        err_path = os.path.join(self.artifact_dir, "stage_APPLICABILITY.mimo.err.log")
@@ -1136,7 +1139,8 @@ class StageExecutor:
        prompt = chr(10).join(pp)

        mimo_bin = os.environ.get("MIMO_BIN", "/var/lib/gitea-desktop-runner/bin/mimo")
-        mimo_model = os.environ.get("MIMO_MODEL", "Verboo Code/glm-5.3-flash")
+        from rune.provider_context import require_model
+        mimo_model = require_model()  # T16 §16F: provider layer owns selection
        mimo_agent = os.environ.get("MIMO_AGENT", "build")
        events_path = os.path.join(self.artifact_dir, "stage_ADVERSARIAL.events.jsonl")
        err_path = os.path.join(self.artifact_dir, "stage_ADVERSARIAL.mimo.err.log")
@@ -1344,7 +1348,8 @@ class StageExecutor:
        prompt = chr(10).join(pp).replace("{n}", str(self.issue))

        mimo_bin = os.environ.get("MIMO_BIN", "/var/lib/gitea-desktop-runner/bin/mimo")
-        mimo_model = os.environ.get("MIMO_MODEL", "Verboo Code/glm-5.3-flash")
+        from rune.provider_context import require_model
+        mimo_model = require_model()  # T16 §16F: provider layer owns selection
        mimo_agent = os.environ.get("MIMO_AGENT", "build")
        events_path = os.path.join(self.artifact_dir, "stage_INNOVATION.events.jsonl")
        err_path = os.path.join(self.artifact_dir, "stage_INNOVATION.mimo.err.log")
@@ -1533,7 +1538,8 @@ class StageExecutor:
        prompt = chr(10).join(pp)

        mimo_bin = os.environ.get("MIMO_BIN", "/var/lib/gitea-desktop-runner/bin/mimo")
-        mimo_model = os.environ.get("MIMO_MODEL", "Verboo Code/glm-5.3-flash")
+        from rune.provider_context import require_model
+        mimo_model = require_model()  # T16 §16F: provider layer owns selection
        mimo_agent = os.environ.get("MIMO_AGENT", "build")
        events_path = os.path.join(self.artifact_dir, "stage_SYNTHESIS.events.jsonl")
        err_path = os.path.join(self.artifact_dir, "stage_SYNTHESIS.mimo.err.log")
@@ -1670,7 +1676,8 @@ class StageExecutor:
            prompt = chr(10).join(pp).replace("{n}", str(self.issue))

            mimo_bin = os.environ.get("MIMO_BIN", "/var/lib/gitea-desktop-runner/bin/mimo")
-            mimo_model = os.environ.get("MIMO_MODEL", "Verboo Code/glm-5.3-flash")
+            from rune.provider_context import require_model
+            mimo_model = require_model()  # T16 §16F: provider layer owns selection
            mimo_agent = os.environ.get("MIMO_AGENT", "build")
            events_path = os.path.join(self.artifact_dir, "stage_QUALIFY.events.jsonl")
            err_path = os.path.join(self.artifact_dir, "stage_QUALIFY.mimo.err.log")
@@ -2092,7 +2099,7 @@ class StageExecutor:
    def _fetch_issue(self) -> dict | None:
        import urllib.request

-        token = os.environ.get("GITEA_TOKEN", "")
+        token = gitea_auth.get_token() if gitea_auth.has_token() else ""
        server = os.environ.get("GITEA_SERVER", "https://gitea.thingscouldgetdicey.online")
        base = server.rstrip("/") + "/api/v1/repos/conductor/rune"
        if not token:
@@ -25,6 +25,7 @@ Stage transitions:
  REFINE(unknowns) → targeted return to the earliest deficient stage
    (no full-pipeline restart; INWARD is not rerun unless Rune reality changed)
 """
+
 from __future__ import annotations

 import json
@@ -34,8 +35,14 @@ from pathlib import Path
 from typing import Any

 STAGES = [
-    "ISSUE", "INWARD", "OUTWARD", "APPLICABILITY",
-    "INNOVATION", "ADVERSARIAL", "SYNTHESIS", "QUALIFY",
+    "ISSUE",
+    "INWARD",
+    "OUTWARD",
+    "APPLICABILITY",
+    "INNOVATION",
+    "ADVERSARIAL",
+    "SYNTHESIS",
+    "QUALIFY",
 ]

 # Legacy stage names that must never become READY through the canonical path.
@@ -66,12 +73,15 @@ APPLICABILITY_CLASSES = {
 # The default is SKIP — INNOVATION runs only when warranted.
 APPLICABILITY_ROUTES = {
    "STANDARD_SOLUTION_SUFFICIENT",  # → ADVERSARIAL (INNOVATION skipped)
-    "INNOVATION_WARRANTED",          # → INNOVATION → ADVERSARIAL
+    "INNOVATION_WARRANTED",  # → INNOVATION → ADVERSARIAL
 }

 # Valid INNOVATION conclusions. NO_INNOVATION_WARRANTED is a SUCCESS result.
 INNOVATION_CONCLUSIONS = {
-    "ADOPT", "ADAPT", "BORROW", "BUILD",
+    "ADOPT",
+    "ADAPT",
+    "BORROW",
+    "BUILD",
    "NO_INNOVATION_WARRANTED",
 }

@@ -80,12 +90,12 @@ INNOVATION_CONCLUSIONS = {
 # Only SURVIVES* dispositions unlock SYNTHESIS. Everything else triggers
 # targeted refinement at the earliest stage that can repair the defect.
 ADVERSARIAL_DISPOSITIONS = {
-    "SURVIVES",                    # → SYNTHESIS READY
-    "SURVIVES_WITH_MITIGATIONS",   # → SYNTHESIS READY (mitigations are required inputs)
-    "EVIDENCE_DEFICIENT",          # → targeted OUTWARD → APPLICABILITY → ADVERSARIAL
-    "APPLICABILITY_DEFICIENT",     # → APPLICABILITY → ADVERSARIAL
-    "INNOVATION_WARRANTED",        # → INNOVATION → ADVERSARIAL
-    "NO_VIABLE_CANDIDATE",         # → targeted OUTWARD (expand candidate space)
+    "SURVIVES",  # → SYNTHESIS READY
+    "SURVIVES_WITH_MITIGATIONS",  # → SYNTHESIS READY (mitigations are required inputs)
+    "EVIDENCE_DEFICIENT",  # → targeted OUTWARD → APPLICABILITY → ADVERSARIAL
+    "APPLICABILITY_DEFICIENT",  # → APPLICABILITY → ADVERSARIAL
+    "INNOVATION_WARRANTED",  # → INNOVATION → ADVERSARIAL
+    "NO_VIABLE_CANDIDATE",  # → targeted OUTWARD (expand candidate space)
 }
 SYNTHESIS_UNLOCKING_DISPOSITIONS = {"SURVIVES", "SURVIVES_WITH_MITIGATIONS"}

@@ -99,9 +109,9 @@ ADVERSARIAL_REFINEMENT_ROUTES = {

 # QUALIFY outcomes (narrowed semantics, §8–10 of the correction).
 QUALIFY_OUTCOMES = {
-    "PROCEED",                # implementation-ready → MiMo Code
-    "REFINE_SYNTHESIS",       # build instructions weak → SYNTHESIS only
-    "REOPEN_EARLIER_STAGE",   # exceptional: genuinely new blocking fact
+    "PROCEED",  # implementation-ready → MiMo Code
+    "REFINE_SYNTHESIS",  # build instructions weak → SYNTHESIS only
+    "REOPEN_EARLIER_STAGE",  # exceptional: genuinely new blocking fact
 }


@@ -184,10 +194,14 @@ class ResearchTaskTree:
        if stage not in self.state["completed_stages"] or stage != "QUALIFY":
            self.state["completed_stages"].append(stage)
        self.state["artifacts"][stage] = artifact
-        self.state["stage_history"].append({
-            "stage": stage, "round": self.round,
-            "completed_at": "auto", "artifact_keys": list(artifact.keys()),
-        })
+        self.state["stage_history"].append(
+            {
+                "stage": stage,
+                "round": self.round,
+                "completed_at": "auto",
+                "artifact_keys": list(artifact.keys()),
+            }
+        )
        # Check if we're in a targeted branch
        targeted = self.state.get("targeted_branch_stages", [])
        if targeted and stage in targeted:
@@ -239,9 +253,7 @@ class ResearchTaskTree:
        if disposition in SYNTHESIS_UNLOCKING_DISPOSITIONS:
            ok, problems = self.synthesis_readiness(artifact)
            if not ok:
-                raise ValueError(
-                    "SYNTHESIS readiness invariant failed: " + "; ".join(problems)
-                )
+                raise ValueError("SYNTHESIS readiness invariant failed: " + "; ".join(problems))
            return "SYNTHESIS"
        return ADVERSARIAL_REFINEMENT_ROUTES[disposition]

@@ -258,9 +270,7 @@ class ResearchTaskTree:
        """
        outcome = artifact.get("outcome", "PROCEED")
        if outcome not in QUALIFY_OUTCOMES:
-            raise ValueError(
-                f"QUALIFY outcome must be one of {sorted(QUALIFY_OUTCOMES)}, got: {outcome!r}"
-            )
+            raise ValueError(f"QUALIFY outcome must be one of {sorted(QUALIFY_OUTCOMES)}, got: {outcome!r}")
        self.state["qualify_outcome"] = outcome
        if outcome == "PROCEED":
            self.state["status"] = "COMPLETE"
@@ -268,13 +278,15 @@ class ResearchTaskTree:
        if outcome == "REFINE_SYNTHESIS":
            return "SYNTHESIS"
        # REOPEN_EARLIER_STAGE: all four required fields must be present
-        required_fields = ("reason", "return_to", "new_blocking_fact",
-                           "why_previous_gate_could_not_reasonably_have_caught_it")
+        required_fields = (
+            "reason",
+            "return_to",
+            "new_blocking_fact",
+            "why_previous_gate_could_not_reasonably_have_caught_it",
+        )
        missing = [f for f in required_fields if not artifact.get(f)]
        if missing:
-            raise ValueError(
-                f"REOPEN_EARLIER_STAGE requires {required_fields}; missing: {missing}"
-            )
+            raise ValueError(f"REOPEN_EARLIER_STAGE requires {required_fields}; missing: {missing}")
        return_to = artifact["return_to"]
        if return_to not in STAGES or return_to == "QUALIFY":
            raise ValueError(f"REOPEN_EARLIER_STAGE return_to must be an epistemic stage, got: {return_to!r}")
@@ -298,11 +310,12 @@ class ResearchTaskTree:
        if not survivors:
            problems.append("no surviving candidate")
        if disposition == "SURVIVES_WITH_MITIGATIONS" and not adversarial_artifact.get("required_mitigations"):
-            problems.append("SURVIVES_WITH_MITIGATIONS without required_mitigations — "
-                            "mitigations are mandatory synthesis inputs, not optional prose")
+            problems.append(
+                "SURVIVES_WITH_MITIGATIONS without required_mitigations — "
+                "mitigations are mandatory synthesis inputs, not optional prose"
+            )
        if self.state.get("blocking_unknowns"):
-            problems.append("blocking adversarial unknowns present: "
-                            f"{len(self.state['blocking_unknowns'])}")
+            problems.append(f"blocking adversarial unknowns present: {len(self.state['blocking_unknowns'])}")
        return (len(problems) == 0, problems)

    def _route_after_applicability(self, artifact: dict) -> str:
@@ -320,17 +333,20 @@ class ResearchTaskTree:
        route = artifact.get("route", "STANDARD_SOLUTION_SUFFICIENT")
        if route not in ("STANDARD_SOLUTION_SUFFICIENT", "INNOVATION_WARRANTED"):
            raise ValueError(
-                f"APPLICABILITY route must be STANDARD_SOLUTION_SUFFICIENT or "
-                f"INNOVATION_WARRANTED, got: {route!r}"
+                f"APPLICABILITY route must be STANDARD_SOLUTION_SUFFICIENT or INNOVATION_WARRANTED, got: {route!r}"
            )
        self.state["applicability_route"] = route
        return "INNOVATION" if route == "INNOVATION_WARRANTED" else "ADVERSARIAL"

    def fail(self, reason: str) -> None:
-        self.state["stage_history"].append({
-            "stage": self.state["current_stage"], "round": self.round,
-            "failed": True, "reason": reason,
-        })
+        self.state["stage_history"].append(
+            {
+                "stage": self.state["current_stage"],
+                "round": self.round,
+                "failed": True,
+                "reason": reason,
+            }
+        )
        self.save()

    def more_research(self, unknowns: list[dict]) -> list[dict]:
@@ -360,11 +376,13 @@ class ResearchTaskTree:
                stages_needed = ["INNOVATION"]
            else:
                stages_needed = ["OUTWARD"]
-            branches.append({
-                "unknown": u,
-                "stages": stages_needed,
-                "round": self.round,
-            })
+            branches.append(
+                {
+                    "unknown": u,
+                    "stages": stages_needed,
+                    "round": self.round,
+                }
+            )
            all_stages.extend(stages_needed)

        # Route to the earliest deficient stage in canonical order,
@@ -380,8 +398,7 @@ class ResearchTaskTree:
    def get_prior_artifacts(self, for_stage: str) -> dict:
        """Return the accepted artifacts this stage needs."""
        idx = STAGES.index(for_stage) if for_stage in STAGES else len(STAGES)
-        return {STAGES[i]: self.state["artifacts"].get(STAGES[i], {})
-                for i in range(idx)}
+        return {STAGES[i]: self.state["artifacts"].get(STAGES[i], {}) for i in range(idx)}

    def capsule(self) -> dict:
        """The process capsule for MiMo invocation (§14)."""
@@ -394,11 +411,23 @@ class ResearchTaskTree:
            "accepted_artifacts": {
                # Adversarial-gated refinement (correction §5–6): mitigations
                # and falsification tests MUST reach SYNTHESIS as required inputs.
-                k: {kk: v for kk, v in vv.items() if kk in (
-                    "decision", "claims", "sources", "candidates", "disposition",
-                    "required_mitigations", "falsification_tests",
-                    "surviving_candidates", "rejected_candidates", "route",
-                )}
+                k: {
+                    kk: v
+                    for kk, v in vv.items()
+                    if kk
+                    in (
+                        "decision",
+                        "claims",
+                        "sources",
+                        "candidates",
+                        "disposition",
+                        "required_mitigations",
+                        "falsification_tests",
+                        "surviving_candidates",
+                        "rejected_candidates",
+                        "route",
+                    )
+                }
                for k, vv in self.state["artifacts"].items()
            },
            "blocking_unknowns": self.state.get("blocking_unknowns", []),
@@ -0,0 +1,259 @@
+#!/usr/bin/env python3
+"""rollback_generation.py — T21.4 §21.4F: the governed rollback API.
+
+    rollback_generation(target_generation, expected_current, reason)
+
+Prior tranches proved rollback by recreating deployment primitives by hand
+(ln -sfn against the live deploy root). That is operator work, not governed
+machinery — it bypasses lock, eligibility, health, and history. This module
+is the single canonical rollback operation, built ON the proven deploy
+machinery (deploy_promotion's lock + atomic swap + build-identity authority),
+not beside it.
+
+The promotion authority (deploy_promotion.py) is deliberately MONOTONIC: an
+ancestor promotion is refused as STALE_SUPERSEDED. Regression is therefore
+impossible through promotion — and legitimate only through THIS operation,
+which owns its own obligations:
+
+ 1.  deployment lock       — the SAME .promotion.lock as promote (serialize)
+ 2.  current verification  — build-identity readback must equal expected_current
+ 3.  target existence      — generation dir present, build identity readable
+ 4.  eligibility           — target commit is ancestor-or-equal of current
+                             (regression only; forward = promotion's job;
+                             divergent/unknown refused)
+ 5.  atomic switch         — rename(2) symlink swap, never absent/half-updated
+ 6.  runtime handoff       — Rune executes from current/src at process start;
+                             handoff = identity a fresh launch would run
+ 7.  runtime verification  — fresh-launch identity must equal target
+ 8.  health qualification  — import smoke against target generation src
+ 9.  durable history       — JSONL rollback events, appended inside the lock
+ 10. failure recovery      — health failure restores prior current (where safe)
+ 11. coherent last-good    — last-good := the generation that WAS current
+                             (same invariant promotion maintains)
+
+Truthful outcomes:
+    ROLLED_BACK              target active, health green, history recorded
+    NOOP_ALREADY_CURRENT     target already the deployed generation
+    ROLLED_BACK_REVERTED     switched, health red, prior current restored
+    REFUSED_UNKNOWN_CURRENT  deployed identity unreadable (fail closed)
+    REFUSED_EXPECTED_MISMATCH current != expected_current (stale caller)
+    REFUSED_TARGET_MISSING   target generation dir absent
+    REFUSED_TARGET_UNREADABLE target build identity unreadable
+    REFUSED_NOT_ELIGIBLE     divergent, newer, or non-ancestor target
+    REFUSED_NO_REASON        empty reason (accountability is mandatory)
+"""
+
+from __future__ import annotations
+
+import fcntl
+import json
+import subprocess
+import sys
+import time
+from pathlib import Path
+
+_SCRIPT_DIR = Path(__file__).resolve().parent
+if str(_SCRIPT_DIR) not in sys.path:
+    sys.path.insert(0, str(_SCRIPT_DIR))
+import deploy_promotion  # proven machinery: lock path, ancestry, swaps
+
+DEPLOY_ROOT = deploy_promotion.DEPLOY_ROOT
+HARNESS_ROOT = _SCRIPT_DIR.parent
+HISTORY_PATH = HARNESS_ROOT / "store" / "scaffold" / "rollback_history.jsonl"
+
+ROLLED_BACK = "ROLLED_BACK"
+NOOP = "NOOP_ALREADY_CURRENT"
+ROLLED_BACK_REVERTED = "ROLLED_BACK_REVERTED"
+REFUSED_UNKNOWN_CURRENT = "REFUSED_UNKNOWN_CURRENT"
+REFUSED_EXPECTED_MISMATCH = "REFUSED_EXPECTED_MISMATCH"
+REFUSED_TARGET_MISSING = "REFUSED_TARGET_MISSING"
+REFUSED_TARGET_UNREADABLE = "REFUSED_TARGET_UNREADABLE"
+REFUSED_NOT_ELIGIBLE = "REFUSED_NOT_ELIGIBLE"
+REFUSED_NO_REASON = "REFUSED_NO_REASON"
+
+
+def _identity(deploy_root: Path, generation: str) -> str:
+    """Authoritative commit for a generation dir (build identity, not name)."""
+    gen_dir = deploy_root / generation
+    identity = gen_dir / "src" / "rune" / "_build_identity.json"
+    try:
+        data = json.loads(identity.read_text(encoding="utf-8"))
+        return str(data.get("commit", ""))
+    except (OSError, ValueError):
+        return ""
+
+
+def _health(deploy_root: Path, generation: str, timeout: int = 60) -> tuple[bool, str]:
+    """Health qualification for a generation: build identity readable AND
+    `import rune` succeeds from that generation's src tree."""
+    gen_dir = deploy_root / generation
+    commit = _identity(deploy_root, generation)
+    if not commit:
+        return False, "build identity unreadable"
+    src = gen_dir / "src"
+    try:
+        proc = subprocess.run(
+            [sys.executable, "-c", "import rune; import sys; "
+             "sys.stdout.write(getattr(rune, '__version__', 'ok'))"],
+            env={"PYTHONPATH": str(src), "PATH": "/usr/bin:/bin",
+                 "HOME": str(gen_dir)},
+            capture_output=True, text=True, timeout=timeout, cwd=str(gen_dir),
+        )
+    except subprocess.TimeoutExpired:
+        return False, f"import smoke timed out after {timeout}s"
+    if proc.returncode != 0:
+        return False, f"import smoke failed: {(proc.stderr or '')[-200:]}"
+    return True, f"identity={commit[:12]} import_smoke=ok ({proc.stdout.strip()[:40]})"
+
+
+def _append_history(record: dict, history_path: Path = HISTORY_PATH) -> None:
+    history_path.parent.mkdir(parents=True, exist_ok=True)
+    with open(history_path, "a", encoding="utf-8") as f:
+        f.write(json.dumps(record, sort_keys=True) + "\n")
+
+
+def _swap(deploy_root: Path, name: str, target: Path) -> None:
+    deploy_promotion._atomic_symlink_target(deploy_root, name, target)
+
+
+def rollback_generation(
+    target_generation: str,
+    expected_current: str,
+    reason: str,
+    *,
+    deploy_root: Path = DEPLOY_ROOT,
+    repo: Path = HARNESS_ROOT,
+    history_path: Path = HISTORY_PATH,
+) -> tuple[str, dict]:
+    """The governed rollback operation. Returns (outcome, detail).
+
+    Refusals happen BEFORE any mutation of current/last-good. The lock is
+    shared with promote() so rollback and promotion can never interleave.
+    """
+    if not reason or not reason.strip():
+        return REFUSED_NO_REASON, {"error": "rollback requires a reason"}
+
+    lock_path = deploy_root / ".promotion.lock"
+    # The lock file may be owned by the CI runner (promotion side opens it
+    # for writing). flock needs only an open fd, never write permission, so
+    # open read-only when write is denied — same file, same fence.
+    try:
+        lock_fd = open(lock_path, "a+")
+    except PermissionError:
+        lock_fd = open(lock_path) if lock_path.exists() else \
+            open(lock_path, "a+")
+    with lock_fd:
+        fcntl.flock(lock_fd, fcntl.LOCK_EX)
+        try:
+            current = deploy_root / "current"
+            last_good = deploy_root / "last-good"
+            detail: dict = {"target": target_generation,
+                            "expected_current": expected_current,
+                            "reason": reason[:200],
+                            "at": time.strftime("%Y-%m-%dT%H:%M:%SZ",
+                                                time.gmtime())}
+
+            # 2. current verification — identity authority, never the name
+            current_generation = ""
+            try:
+                current_generation = Path(os_readlink(current)).name
+            except OSError:
+                pass
+            deployed_commit = deploy_promotion.read_deployed_commit(deploy_root)
+            if not deployed_commit:
+                return REFUSED_UNKNOWN_CURRENT, {
+                    "error": "deployed commit unreadable — fail closed",
+                    **detail}
+            detail["current_generation"] = current_generation
+            detail["current_commit"] = deployed_commit
+            if expected_current and expected_current != deployed_commit:
+                detail["error"] = (
+                    f"current is {deployed_commit[:12]}, "
+                    f"caller expected {expected_current[:12]} — refused "
+                    "before mutation")
+                return REFUSED_EXPECTED_MISMATCH, detail
+
+            # 3. target existence + identity
+            target_dir = deploy_root / target_generation
+            if not target_dir.is_dir():
+                detail["error"] = f"generation dir {target_generation} absent"
+                return REFUSED_TARGET_MISSING, detail
+            target_commit = _identity(deploy_root, target_generation)
+            if not target_commit:
+                detail["error"] = (
+                    f"generation {target_generation} build identity unreadable")
+                return REFUSED_TARGET_UNREADABLE, detail
+            detail["target_commit"] = target_commit
+
+            if target_commit == deployed_commit:
+                detail["note"] = "target already the deployed generation"
+                return NOOP, detail
+
+            # 4. eligibility — ancestor-or-equal only (regression)
+            rel = deploy_promotion.ancestry_of(target_commit, deployed_commit,
+                                               repo)
+            if rel != "ancestor":
+                detail["error"] = (
+                    f"target commit ancestry={rel} — only ancestors are "
+                    "rollback-eligible (forward movement belongs to "
+                    "deploy_promotion)")
+                return REFUSED_NOT_ELIGIBLE, detail
+
+            # 5. atomic switch (inside lock), last-good := previous current
+            if last_good.is_symlink():
+                last_good.unlink()
+            _swap(deploy_root, "last-good", Path(os_readlink(current)))
+            _swap(deploy_root, "current", target_dir)
+            detail["switched_to"] = target_generation
+
+            # 7+8. runtime verification + health qualification
+            post_commit = deploy_promotion.read_deployed_commit(deploy_root)
+            health_ok, health_detail = _health(deploy_root, target_generation)
+            detail["post_switch_identity"] = post_commit
+            detail["health"] = health_detail
+
+            record = {"at": detail["at"], "operation": "rollback_generation",
+                      **detail}
+            if post_commit != target_commit or not health_ok:
+                # 10. failure recovery — restore prior current (safe: we
+                # hold the lock and the prior generation is intact)
+                prev = Path(os_readlink(last_good)) if last_good.is_symlink() \
+                    else None
+                if prev is not None:
+                    _swap(deploy_root, "current", prev)
+                detail["reverted_to"] = (Path(str(os_readlink(current))).name
+                                         if current.is_symlink() else "?")
+                detail["outcome"] = ROLLED_BACK_REVERTED
+                record["outcome"] = ROLLED_BACK_REVERTED
+                _append_history(record, history_path)
+                return ROLLED_BACK_REVERTED, detail
+
+            record["outcome"] = ROLLED_BACK
+            _append_history(record, history_path)
+            detail["outcome"] = ROLLED_BACK
+            return ROLLED_BACK, detail
+        finally:
+            fcntl.flock(lock_fd, fcntl.LOCK_UN)
+
+
+def os_readlink(p: Path) -> str:
+    return str(Path(str(p)).readlink()) if hasattr(Path, "readlink") \
+        else str(Path(str(p)).readlink())
+
+
+def main() -> int:
+    import argparse
+    ap = argparse.ArgumentParser(description=__doc__)
+    ap.add_argument("target_generation")
+    ap.add_argument("--expected-current", required=True,
+                    help="commit the caller believes is currently deployed")
+    ap.add_argument("--reason", required=True)
+    args = ap.parse_args()
+    outcome, detail = rollback_generation(args.target_generation,
+                                          args.expected_current, args.reason)
+    print(json.dumps({"outcome": outcome, **detail}, indent=2, sort_keys=True))
+    return 0 if outcome in (ROLLED_BACK, NOOP) else 1
+
+
+if __name__ == "__main__":
+    sys.exit(main())
@@ -14,6 +14,7 @@ This module captures ACTUAL runtime evidence — not configuration inference:
 The evidence record travels with the stage result and is validated by the
 placement gate BEFORE a stage result can be accepted.
 """
+
 from __future__ import annotations

 import os
@@ -30,14 +31,23 @@ DESKTOP_RUNNER_LABEL = "rune-ci-desktop"

 # Stages that require desktop-heavy execution (cognition, research, coding).
 DESKTOP_HEAVY_STAGES = {
-    "INWARD", "OUTWARD", "APPLICABILITY", "INNOVATION",
-    "ADVERSARIAL", "SYNTHESIS", "QUALIFY",
+    "INWARD",
+    "OUTWARD",
+    "APPLICABILITY",
+    "INNOVATION",
+    "ADVERSARIAL",
+    "SYNTHESIS",
+    "QUALIFY",
 }

 # Lightweight forge/control work that MAY legitimately run on the Pi.
 PI_LIGHTWEIGHT_OPERATIONS = {
-    "issue_metadata", "label_change", "artifact_metadata",
-    "process_state", "eligibility_check", "state_transition",
+    "issue_metadata",
+    "label_change",
+    "artifact_metadata",
+    "process_state",
+    "eligibility_check",
+    "state_transition",
 }


@@ -58,14 +68,8 @@ def capture_execution_evidence(
    runner identity present in the environment.
    """
    hostname = socket.gethostname()
-    runner_name = (
-        os.environ.get("RUNNER_NAME")
-        or os.environ.get("GITEA_RUNNER_NAME")
-        or ("local:" + hostname)
-    )
-    runner_labels = [
-        l.strip() for l in os.environ.get("RUNNER_LABELS", "").split(",") if l.strip()
-    ] or [runner_name]
+    runner_name = os.environ.get("RUNNER_NAME") or os.environ.get("GITEA_RUNNER_NAME") or ("local:" + hostname)
+    runner_labels = [l.strip() for l in os.environ.get("RUNNER_LABELS", "").split(",") if l.strip()] or [runner_name]

    return {
        "runner_hostname": hostname,
@@ -92,13 +96,12 @@ def validate_placement(evidence: dict) -> tuple[bool, list[str]]:
    klass = evidence.get("execution_class", "desktop-heavy")

    # Evidence completeness — identity must be actually captured
-    for field in ("runner_hostname", "runner_name", "runner_labels",
-                  "execution_class", "stage", "timestamp"):
+    for field in ("runner_hostname", "runner_name", "runner_labels", "execution_class", "stage", "timestamp"):
        if not evidence.get(field):
            violations.append(f"placement evidence missing field: {field}")

    if klass == "desktop-heavy":
-        identity = (evidence.get("runner_name", "") + " " + evidence.get("runner_hostname", ""))
+        identity = evidence.get("runner_name", "") + " " + evidence.get("runner_hostname", "")
        if _is_pi(identity) or any(_is_pi(l) for l in evidence.get("runner_labels", [])):
            violations.append(
                f"PLACEMENT VIOLATION: desktop-heavy stage '{stage}' executed on Pi "
@@ -109,10 +112,12 @@ def validate_placement(evidence: dict) -> tuple[bool, list[str]]:
        # runner NAME (rune-desktop), or a verified local desktop execution.
        # Pi rejection above already caught Pi markers — this is the allow-side.
        labels = " ".join(evidence.get("runner_labels", []))
-        if (DESKTOP_RUNNER_LABEL not in identity
-                and "rune-desktop" not in identity
-                and DESKTOP_RUNNER_LABEL not in labels
-                and "local:" not in evidence.get("runner_name", "")):
+        if (
+            DESKTOP_RUNNER_LABEL not in identity
+            and "rune-desktop" not in identity
+            and DESKTOP_RUNNER_LABEL not in labels
+            and "local:" not in evidence.get("runner_name", "")
+        ):
            violations.append(
                f"runner '{evidence.get('runner_name')}' is not '{DESKTOP_RUNNER_LABEL}' "
                f"nor the registered rune-desktop runner, and not a verified local "
@@ -0,0 +1,205 @@
+#!/usr/bin/env python3
+"""scaffold_projection.py — Scaffolding V1 Tranche 1 (projection, not authority).
+
+Answers, per component/module:
+  what subsystem owns X?          → SPINE store (structural authority)
+  what does X depend on?          → SPINE relationships (ingested from AST)
+  who consumes X?                 → SPINE relationships (ingested from AST)
+  what changes if X changes?      → mirrors + dependent subsystems + consumers (SPINE)
+  what proof obligations arise?   → unit test, canary, ownership, umbrella gates
+  which obligations are complete? → evidence check (test file exists, canary registered)
+
+SPINE INTEGRATION: structural facts come from the Spine store — the single
+structural authority (ingested once from the umbrella manifest + AST).
+This script adds change-impact and proof-obligation reasoning on top; it
+no longer re-derives structure. No store → builds one (deterministic).
+
+Commands:
+  component <module>    full card for one src/rune module
+  impact <module>       expected impact + obligations BEFORE a change
+  obligations <module>  obligation checklist with satisfied/pending state
+  all [outfile]         write the full projection JSON (default docs/)
+"""
+from __future__ import annotations
+import ast, json, sys
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parent.parent
+SRC = ROOT / "src" / "rune"
+TESTS = ROOT / "tests"
+sys.path.insert(0, str(ROOT / "src"))
+from rune import spine  # noqa: E402 — Spine IS the structural authority
+
+
+def load_spine() -> dict:
+    """Structural state from the Spine store (built on first use)."""
+    doc = spine.load()
+    if doc is None:
+        doc = spine.build()
+        spine.save(doc)
+    return doc
+
+
+def module_imports(module: str) -> set[str]:
+    """Module names imported by src/rune/<module>.py (intra-package).
+
+    Includes relative imports (`from .x import y` — the dual-mode lazy
+    pattern used across the runtime; missing these would hide real
+    consumers, observed first run: rune_operator ↔ local_tool_runtime)."""
+    path = SRC / f"{module}.py"
+    if not path.is_file():
+        return set()
+    tree = ast.parse(path.read_text())
+    names: set[str] = set()
+    for node in ast.walk(tree):
+        if isinstance(node, ast.Import):
+            for a in node.names:
+                names.add(a.name.split(".")[0])
+        elif isinstance(node, ast.ImportFrom) and node.module:
+            names.add(node.module.split(".")[0])
+    return names
+
+
+def import_graph() -> dict[str, set[str]]:
+    return {p.stem: module_imports(p.stem) for p in SRC.glob("*.py")}
+
+
+def consumers_of(module: str, graph: dict[str, set[str]]) -> set[str]:
+    return {m for m, deps in graph.items() if module in deps and m != module}
+
+
+def subsystem_for(doc: dict, module: str) -> str:
+    rels = spine.relationships(doc, from_id=f"module:{module}", type="belongs_to")
+    return rels[0]["to"].split(":", 1)[1] if rels else ""
+
+
+def dependent_subsystems(doc: dict, subsystem: str) -> list[str]:
+    return [r["from"].split(":", 1)[1] for r in
+            spine.relationships(doc, to_id=f"subsystem:{subsystem}", type="depends_on")]
+
+
+def mirrors_for(doc: dict, module: str) -> list[str]:
+    subs = spine.relationships(doc, from_id=f"module:{module}", type="belongs_to")
+    out = []
+    for s in subs:
+        out.extend(r["to"].split(":", 1)[1] for r in
+                   spine.relationships(doc, from_id=s["to"], type="mirrors"))
+    return sorted(set(out))
+
+
+def canary_registered(module: str) -> bool:
+    try:
+        sys.path.insert(0, str(SRC))
+        from rune import canary_native
+        return module in canary_native.MODULES
+    except ImportError:
+        # canary_native pulls optional deps (sqlite_vec etc.); the projection
+        # degrades to "not registered" rather than failing the audit.
+        return False
+
+
+def obligations(module: str, doc: dict) -> list[dict]:
+    """Proof obligations for changing this module, with evidence state.
+    Structural facts sourced from Spine."""
+    umb = subsystem_for(doc, module)
+    test_file = TESTS / f"test_{module}.py"
+    obs = [
+        {"obligation": "unit_test", "satisfied": test_file.is_file(),
+         "evidence": f"tests/test_{module}.py" + ("" if test_file.is_file() else " (MISSING)")},
+        {"obligation": "canary", "satisfied": canary_registered(module),
+         "evidence": "canary_native.MODULES" if canary_registered(module)
+                     else "not in canary registry (informational — not every module needs one)"},
+        {"obligation": "umbrella_manifest", "satisfied": bool(umb),
+         "evidence": f"umbrella '{umb}'" if umb else "NOT in any umbrella — manifest gap"},
+        {"obligation": "full_suite", "satisfied": None,
+         "evidence": "run: pytest tests/ -q (runtime evidence, not static)"},
+        {"obligation": "lint_ratchet", "satisfied": None,
+         "evidence": "run: scripts/lint_ratchet.py"},
+        {"obligation": "henxels", "satisfied": None,
+         "evidence": "run: henxels check --all"},
+    ]
+    if umb:
+        deps = dependent_subsystems(doc, umb)
+        if deps:
+            obs.append({"obligation": "dependent_canaries", "satisfied": None,
+                        "evidence": f"subsystems depending on '{umb}': {', '.join(deps)} — their canaries gate the change"})
+        mir = mirrors_for(doc, module)
+        if mir:
+            obs.append({"obligation": "mirrors_travel", "satisfied": None,
+                        "evidence": f"files that must travel together: {', '.join(mir)}"})
+    return obs
+
+
+def component_card(module: str, doc: dict) -> dict:
+    """Structural card from SPINE (owner/deps/consumers/mirrors) + obligations."""
+    deps = [r["to"].split(":", 1)[1] for r in
+            spine.relationships(doc, from_id=f"module:{module}", type="depends_on")]
+    consumers = [r["to"].split(":", 1)[1] for r in
+                 spine.relationships(doc, from_id=f"module:{module}", type="consumed_by")]
+    e = spine.entity(doc, f"module:{module}")
+    card = {
+        "component": module,
+        "subsystem": subsystem_for(doc, module),
+        "dependencies": sorted(set(deps)),
+        "consumers": sorted(set(consumers)),
+        "mirrors": mirrors_for(doc, module),
+        "obligations": obligations(module, doc),
+    }
+    if e:
+        card["spine_maturity"] = e.get("maturity")
+        card["spine_evidence"] = e.get("evidence_refs", [])
+    return card
+
+
+def impact(module: str) -> dict:
+    doc = load_spine()
+    card = component_card(module, doc)
+    umb = card["subsystem"]
+    card["expected_impact"] = {
+        "direct_consumers": card["consumers"],
+        "subsystem_dependents": dependent_subsystems(doc, umb) if umb else [],
+        "must_travel": card["mirrors"],
+        "gates": ["full_suite", "lint_ratchet", "henxels", "umbrella_gates"]
+                + (["canary"] if card and any(o["obligation"] == "canary" and o["satisfied"]
+                                              for o in card["obligations"]) else []),
+    }
+    return card
+
+
+def main() -> int:
+    args = sys.argv[1:]
+    if not args:
+        print(__doc__)
+        return 2
+    cmd = args[0]
+    if cmd in ("component", "impact", "obligations") and len(args) < 2:
+        print(f"usage: {cmd} <module>", file=sys.stderr)
+        return 2
+    doc = load_spine()
+    if cmd == "impact":
+        print(json.dumps(impact(args[1]), indent=1))
+    elif cmd == "component":
+        print(json.dumps(component_card(args[1], doc), indent=1))
+    elif cmd == "obligations":
+        card = component_card(args[1], doc)
+        print(json.dumps(card["obligations"], indent=1))
+    elif cmd == "all":
+        out = Path(args[1]) if len(args) > 1 else ROOT / "docs" / "scaffold_projection.json"
+        cards = {p.stem: component_card(p.stem, doc)
+                 for p in sorted(SRC.glob("*.py"))
+                 if spine.entity(doc, f"module:{p.stem}")}
+        out.write_text(json.dumps({
+            "generated_by": "scripts/scaffold_projection.py (projection — no new authority)",
+            "sources": ["rune.spine store (structural authority)", "src/rune/canary_native.py",
+                        "tests/ filesystem"],
+            "components": cards,
+        }, indent=1) + "\n")
+        print(f"projection written: {out} ({len(cards)} components)")
+    else:
+        print(f"unknown command: {cmd}", file=sys.stderr)
+        return 2
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
@@ -18,6 +18,13 @@ from __future__ import annotations
 import re
 import sys

+# T20 §20B: push-convergence synthesizes PR titles as "<kind>: converged
+# release (push-to-master bump)" where kind comes from
+# version_convergence.highest_warranted_bump. The title text must carry
+# that decision — "minor:" was re-classified as patch here, downgrading
+# every feature tranche (observed: feat ranges bumped 0.66.77-83 not 0.67.0).
+MINOR_PREFIX_RE = re.compile(r"^\s*minor\b", re.IGNORECASE)
+
 BREAKING_RE = re.compile(r"^\s*\w+(\([^)]*\))?!:")
 FEAT_RE = re.compile(r"^\s*feat(\([^)]*\))?")

@@ -25,6 +32,8 @@ FEAT_RE = re.compile(r"^\s*feat(\([^)]*\))?")
 def next_version(version: str, title: str) -> str:
    if BREAKING_RE.match(title) or "BREAKING CHANGE" in title:
        kind = "major"
+    elif MINOR_PREFIX_RE.match(title):
+        kind = "minor"  # synthesized convergence title carries the kind
    elif FEAT_RE.match(title):
        kind = "minor"
    else:
@@ -21,6 +21,7 @@ Hard invariants:
    must not silently falsify the ledger.
  - No trust scores. Factual source classification only.
 """
+
 from __future__ import annotations

 import hashlib
@@ -70,8 +71,7 @@ def build_records(outward: dict, stage: str = "OUTWARD") -> list[dict]:
        url = _norm(r.get("url", ""))
        if not url:
            continue
-        rid = _receipt_id(stage, r.get("tool", "?"), r.get("url", ""),
-                          r.get("ts", 0), i)
+        rid = _receipt_id(stage, r.get("tool", "?"), r.get("url", ""), r.get("ts", 0), i)
        if r.get("kind") == "SOURCE_READ":
            read_urls[url] = {"receipt_id": rid, "tool": r.get("tool", "?"), "ts": r.get("ts")}
        elif r.get("kind") == "TOOL_FAILED":
@@ -93,7 +93,7 @@ def build_records(outward: dict, stage: str = "OUTWARD") -> list[dict]:
    for i, c in enumerate(outward.get("contradictory_evidence", outward.get("contradictions", []))):
        if isinstance(c, dict):
            for sid in c.get("contradicts_source_ids", []):
-                contradictions_by_source.setdefault(sid, []).append(c.get("id", f"C{i+1}"))
+                contradictions_by_source.setdefault(sid, []).append(c.get("id", f"C{i + 1}"))
        # plain-string contradictions are not linked (labeled in render)

    records = []
@@ -119,28 +119,41 @@ def build_records(outward: dict, stage: str = "OUTWARD") -> list[dict]:
        # Upgrade precedence fix is above; this block is intentionally after
        # status resolution so the downgrade applies regardless of note text.
        resolved = s.get("resolved_url", url)  # explicit field if present
-        records.append({
-            "receipt_id": (receipt or {}).get("receipt_id", ""),
-            "source_id": sid,
-            "requested_url": url,
-            "resolved_url": resolved,
-            "domain": _domain(resolved),
-            "status": status,
-            "tool": (receipt or {}).get("tool", s.get("tool", "UNKNOWN_NOT_EXPOSED")),
-            "timestamp": (receipt or {}).get("ts", "UNKNOWN_NOT_EXPOSED"),
-            "source_type": s.get("class", s.get("source_class", "UNKNOWN_NOT_EXPOSED")),
-            "claim_ids": claims_by_source.get(sid, []),
-            "contradiction_ids": contradictions_by_source.get(sid, []),
-            "blocked_reason": (s.get("note", "") if status == STATUS_BLOCKED else
-                               (receipt.get("kind") if status == STATUS_FAILED and receipt else
-                                ("no machine SOURCE_READ receipt matched (legacy artifact)" if status == STATUS_FAILED else ""))),
-            "supports_evidence": status == STATUS_READ,
-        })
+        records.append(
+            {
+                "receipt_id": (receipt or {}).get("receipt_id", ""),
+                "source_id": sid,
+                "requested_url": url,
+                "resolved_url": resolved,
+                "domain": _domain(resolved),
+                "status": status,
+                "tool": (receipt or {}).get("tool", s.get("tool", "UNKNOWN_NOT_EXPOSED")),
+                "timestamp": (receipt or {}).get("ts", "UNKNOWN_NOT_EXPOSED"),
+                "source_type": s.get("class", s.get("source_class", "UNKNOWN_NOT_EXPOSED")),
+                "claim_ids": claims_by_source.get(sid, []),
+                "contradiction_ids": contradictions_by_source.get(sid, []),
+                "blocked_reason": (
+                    s.get("note", "")
+                    if status == STATUS_BLOCKED
+                    else (
+                        receipt.get("kind")
+                        if status == STATUS_FAILED and receipt
+                        else (
+                            "no machine SOURCE_READ receipt matched (legacy artifact)"
+                            if status == STATUS_FAILED
+                            else ""
+                        )
+                    )
+                ),
+                "supports_evidence": status == STATUS_READ,
+            }
+        )
    return records


-def render_markdown(records: list[dict], outward: dict, runner: dict | None,
-                    model: str, artifact_hash: str, legacy_claims: bool = False) -> str:
+def render_markdown(
+    records: list[dict], outward: dict, runner: dict | None, model: str, artifact_hash: str, legacy_claims: bool = False
+) -> str:
    """Fixed Markdown template. Pure function of records + metadata."""
    d = outward.get("discovery", {})
    n_read = sum(1 for r in records if r["status"] == STATUS_READ)
@@ -152,9 +165,11 @@ def render_markdown(records: list[dict], outward: dict, runner: dict | None,
    runner_host = (runner or {}).get("runner_hostname", "UNKNOWN")
    prov_note = ""
    if legacy_claims:
-        prov_note = ("\n*Claim↔source links for this legacy artifact are derived from "
-                     "claim nesting (pre-#282 format); future artifacts carry explicit "
-                     "supports_source_ids/contradicts_source_ids.*")
+        prov_note = (
+            "\n*Claim↔source links for this legacy artifact are derived from "
+            "claim nesting (pre-#282 format); future artifacts carry explicit "
+            "supports_source_ids/contradicts_source_ids.*"
+        )

    lines = [
        "## OUTWARD — ACCEPTED",
@@ -179,6 +194,7 @@ def render_markdown(records: list[dict], outward: dict, runner: dict | None,
        link = f"[{r['domain']}]({r['requested_url']})" if r["requested_url"].startswith("http") else r["domain"]
        lines.append(
            f"| {r['source_id']} | {link} | {r['source_type']} | {r['status']} "
-            f"| {len(r['claim_ids'])} | {len(r['contradiction_ids'])} | {caveat[:60]} |")
+            f"| {len(r['claim_ids'])} | {len(r['contradiction_ids'])} | {caveat[:60]} |"
+        )
    lines.append(prov_note)
    return "\n".join(lines)
@@ -13,68 +13,102 @@ contract, the contract-drift test fails.
 Compatibility aliases (e.g. urls → url, classification → source_class)
 are explicit, versioned, and bounded — not arbitrary rescue behavior.
 """
+
 from __future__ import annotations

 # Per-stage contract: required fields in the structured artifact
 STAGE_CONTRACTS: dict[str, dict] = {
    "INWARD": {
        "required_fields": [
-            {"name": "claims", "type": "list",
-             "description": "relevant files/behaviors with evidence citations"},
+            {"name": "claims", "type": "list", "description": "relevant files/behaviors with evidence citations"},
        ],
        "optional_fields": ["context", "deterministic"],
    },
    "OUTWARD": {
        "required_fields": [
-            {"name": "discovery", "type": "object",
-             "description": "queries, sources_discovered, source_count"},
-            {"name": "reading", "type": "object",
-             "description": "sources_selected, sources_read (with machine receipts), evidence_extracted"},
-            {"name": "tool_receipts", "type": "list",
-             "description": "per-event receipts with kind/tool/url/timestamp"},
-            {"name": "coverage", "type": "object",
-             "description": "required_classes, represented_classes, uncovered"},
+            {"name": "discovery", "type": "object", "description": "queries, sources_discovered, source_count"},
+            {
+                "name": "reading",
+                "type": "object",
+                "description": "sources_selected, sources_read (with machine receipts), evidence_extracted",
+            },
+            {"name": "tool_receipts", "type": "list", "description": "per-event receipts with kind/tool/url/timestamp"},
+            {"name": "coverage", "type": "object", "description": "required_classes, represented_classes, uncovered"},
+        ],
+        "optional_fields": [
+            "text",
+            "limitations",
+            "contradictory_evidence",
+            "supporting_evidence",
+            "alternatives",
+            "adjacent_field_leads",
+            "unresolved_questions",
        ],
-        "optional_fields": ["text", "limitations", "contradictory_evidence",
-                            "supporting_evidence", "alternatives",
-                            "adjacent_field_leads", "unresolved_questions"],
    },
    "APPLICABILITY": {
        "required_fields": [
-            {"name": "route", "type": "string",
-             "description": "STANDARD_SOLUTION_SUFFICIENT or INNOVATION_WARRANTED",
-             "enum": ["STANDARD_SOLUTION_SUFFICIENT", "INNOVATION_WARRANTED"]},
-            {"name": "candidates", "type": "list",
-             "description": "classified mechanisms with integration_point, evidence, source_ids"},
+            {
+                "name": "route",
+                "type": "string",
+                "description": "STANDARD_SOLUTION_SUFFICIENT or INNOVATION_WARRANTED",
+                "enum": ["STANDARD_SOLUTION_SUFFICIENT", "INNOVATION_WARRANTED"],
+            },
+            {
+                "name": "candidates",
+                "type": "list",
+                "description": "classified mechanisms with integration_point, evidence, source_ids",
+            },
        ],
        "optional_fields": ["notes"],
    },
    "ADVERSARIAL": {
        "required_fields": [
-            {"name": "disposition", "type": "string",
-             "description": "SURVIVES | SURVIVES_WITH_MITIGATIONS | EVIDENCE_DEFICIENT | "
-                            "APPLICABILITY_DEFICIENT | INNOVATION_WARRANTED | NO_VIABLE_CANDIDATE",
-             "enum": ["SURVIVES", "SURVIVES_WITH_MITIGATIONS", "EVIDENCE_DEFICIENT",
-                      "APPLICABILITY_DEFICIENT", "INNOVATION_WARRANTED", "NO_VIABLE_CANDIDATE"]},
-            {"name": "surviving_candidates", "type": "list",
-             "description": "candidate ids that survived attack"},
-            {"name": "attacks", "type": "list",
-             "description": "attack rows with target_candidate, dimension, finding, severity"},
-            {"name": "required_mitigations", "type": "list",
-             "description": "mitigations that must appear in SYNTHESIS (mandatory, not prose)"},
-            {"name": "falsification_tests", "type": "list",
-             "description": "kill-tests that must convert into invariants"},
+            {
+                "name": "disposition",
+                "type": "string",
+                "description": "SURVIVES | SURVIVES_WITH_MITIGATIONS | EVIDENCE_DEFICIENT | "
+                "APPLICABILITY_DEFICIENT | INNOVATION_WARRANTED | NO_VIABLE_CANDIDATE",
+                "enum": [
+                    "SURVIVES",
+                    "SURVIVES_WITH_MITIGATIONS",
+                    "EVIDENCE_DEFICIENT",
+                    "APPLICABILITY_DEFICIENT",
+                    "INNOVATION_WARRANTED",
+                    "NO_VIABLE_CANDIDATE",
+                ],
+            },
+            {"name": "surviving_candidates", "type": "list", "description": "candidate ids that survived attack"},
+            {
+                "name": "attacks",
+                "type": "list",
+                "description": "attack rows with target_candidate, dimension, finding, severity",
+            },
+            {
+                "name": "required_mitigations",
+                "type": "list",
+                "description": "mitigations that must appear in SYNTHESIS (mandatory, not prose)",
+            },
+            {
+                "name": "falsification_tests",
+                "type": "list",
+                "description": "kill-tests that must convert into invariants",
+            },
        ],
-        "optional_fields": ["rejected_candidates", "critical_assumptions",
-                            "new_research_gaps"],
+        "optional_fields": ["rejected_candidates", "critical_assumptions", "new_research_gaps"],
    },
    "INNOVATION": {
        "required_fields": [
-            {"name": "conclusion", "type": "string",
-             "description": "ADOPT | ADAPT | BORROW | BUILD | NO_INNOVATION_WARRANTED",
-             "enum": ["ADOPT", "ADAPT", "BORROW", "BUILD", "NO_INNOVATION_WARRANTED"]},
-            {"name": "candidates", "type": "list",
-             "description": "each with derived_from (mandatory), kill_test (mandatory)"},
+            {
+                "name": "conclusion",
+                "type": "string",
+                "description": "ADOPT | ADAPT | BORROW | BUILD | NO_INNOVATION_WARRANTED",
+                "enum": ["ADOPT", "ADAPT", "BORROW", "BUILD", "NO_INNOVATION_WARRANTED"],
+            },
+            {
+                "name": "candidates",
+                "type": "list",
+                "description": "each with derived_from (mandatory), kill_test (mandatory)",
+            },
        ],
        "optional_fields": ["notes"],
    },
@@ -83,23 +117,33 @@ STAGE_CONTRACTS: dict[str, dict] = {
            {"name": "selected_design", "type": "string"},
            {"name": "implementation_sequence", "type": "list"},
            {"name": "interfaces", "type": "list"},
-            {"name": "invariants", "type": "list",
-             "description": "every falsification test must be converted into an invariant"},
+            {
+                "name": "invariants",
+                "type": "list",
+                "description": "every falsification test must be converted into an invariant",
+            },
            {"name": "tests_required", "type": "list"},
            {"name": "acceptance_criteria", "type": "list"},
            {"name": "rollback", "type": "string"},
-            {"name": "data_flow", "type": "list",
-             "description": "how data moves through the designed system"},
+            {"name": "data_flow", "type": "list", "description": "how data moves through the designed system"},
+        ],
+        "optional_fields": [
+            "why_this_design_won",
+            "rejected_alternatives",
+            "components_affected",
+            "authority_boundaries",
+            "failure_modes",
+            "deliberately_excluded",
        ],
-        "optional_fields": ["why_this_design_won", "rejected_alternatives",
-                            "components_affected", "authority_boundaries",
-                            "failure_modes", "deliberately_excluded"],
    },
    "QUALIFY": {
        "required_fields": [
-            {"name": "outcome", "type": "string",
-             "description": "PROCEED | REFINE_SYNTHESIS | REOPEN_EARLIER_STAGE",
-             "enum": ["PROCEED", "REFINE_SYNTHESIS", "REOPEN_EARLIER_STAGE"]},
+            {
+                "name": "outcome",
+                "type": "string",
+                "description": "PROCEED | REFINE_SYNTHESIS | REOPEN_EARLIER_STAGE",
+                "enum": ["PROCEED", "REFINE_SYNTHESIS", "REOPEN_EARLIER_STAGE"],
+            },
        ],
        "optional_fields": ["findings", "deterministic_checks", "reopen"],
    },
@@ -109,7 +153,7 @@ STAGE_CONTRACTS: dict[str, dict] = {
 # (bounded, versioned, NOT arbitrary rescue behavior).
 COMPATIBILITY_ALIASES: dict[str, dict[str, str]] = {
    "OUTWARD": {
-        "urls": "url",              # worker uses urls (list) → take first as url
+        "urls": "url",  # worker uses urls (list) → take first as url
        "classification": "source_class",
        "search_receipts": "queries",
        "read_receipts": "sources_read",
@@ -27,6 +27,7 @@ Attempt lifecycle (from the executor's perspective):
                                        TIMEOUT | INTERRUPTED
  persist(handle)                     → append + mirror
 """
+
 from __future__ import annotations

 import json
@@ -45,8 +46,7 @@ def _now_iso() -> str:
 class StageAttempt:
    """One attempt's observed lifecycle. Pure data — no control flow."""

-    def __init__(self, attempt_id: str, event_id: str, stage: str,
-                 issue: int, process_id: str):
+    def __init__(self, attempt_id: str, event_id: str, stage: str, issue: int, process_id: str):
        self.attempt_id = attempt_id
        self.event_id = event_id
        self.stage = stage
@@ -121,8 +121,7 @@ class StageAttempt:
        self.terminal_status = terminal_status
        self.finished_at = _now_iso()
        # Close any unclosed windows (INTERRUPTED/TIMEOUT paths)
-        for lst in (self.model_windows, self.tool_windows,
-                    self.gate_windows, self.persistence_windows):
+        for lst in (self.model_windows, self.tool_windows, self.gate_windows, self.persistence_windows):
            if lst and lst[-1][1] == 0.0:
                s, _ = lst[-1]
                lst[-1] = (s, time.monotonic())
@@ -160,7 +159,8 @@ class StageAttempt:
            # not double-counted work.
            "overlap_note": (
                "categories are observed windows and may overlap; "
-                "wall_time is authoritative, category sums are not additive"),
+                "wall_time is authoritative, category sums are not additive"
+            ),
            "model_calls": self.model_calls,
            "tool_calls": self.tool_calls,
            "repo_read_calls": self.repo_read_calls,
@@ -0,0 +1,64 @@
+#!/usr/bin/env python3
+"""start_wr_daemon.py — T21.5 §7: THE canonical daemon restart launcher.
+
+T21.4 live finding: `python3 daemon.py` defaults WR_ROOT to the daemon's
+install dir, so a restart silently opened a second event store. This
+launcher resolves the root ONCE through scripts/wr_root_resolver.py (the
+single authority) and passes it explicitly — nobody has to remember to
+export WR_ROOT.
+"""
+from __future__ import annotations
+
+import os
+import subprocess
+import sys
+import time
+import urllib.request
+from pathlib import Path
+
+_SCRIPT_DIR = Path(__file__).resolve().parent
+sys.path.insert(0, str(_SCRIPT_DIR))
+import wr_root_resolver as resolver  # noqa: E402
+
+DAEMON_CODE_CANDIDATES = (
+    "/home/nick/.local/share/mimo-wait-resume/daemon.py",
+)
+
+
+def find_daemon_code() -> Path:
+    for cand in DAEMON_CODE_CANDIDATES:
+        p = Path(cand)
+        if p.is_file():
+            return p
+    raise RuntimeError(f"daemon.py not found in {DAEMON_CODE_CANDIDATES}")
+
+
+def main() -> int:
+    root = resolver.resolve_root()
+    daemon = find_daemon_code()
+    env = dict(os.environ,
+               WR_ROOT=str(root),            # daemon data root
+               RUNE_WR_ROOT=str(root),       # adapter-side authority
+               RUNE_WR_CODE=str(daemon.parent))
+    port = sys.argv[1] if len(sys.argv) > 1 else "8791"
+    proc = subprocess.Popen(
+        [sys.executable, str(daemon), port],
+        env=env, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL,
+        start_new_session=True)
+    # bounded startup probe (process-local, not an external wait)
+    for _ in range(20):
+        try:
+            with urllib.request.urlopen(
+                    f"http://127.0.0.1:{port}/health", timeout=1) as r:
+                if r.status == 200:
+                    print(f"wr daemon up: pid={proc.pid} root={root} "
+                          f"port={port}")
+                    return 0
+        except OSError:
+            time.sleep(0.25)
+    print(f"wr daemon did not become healthy (pid={proc.pid})", file=sys.stderr)
+    return 1
+
+
+if __name__ == "__main__":
+    sys.exit(main())
@@ -0,0 +1,331 @@
+#!/usr/bin/env python3
+"""structural_coverage_audit.py — T21.4 §21.4G: Rune-wide meaningful
+structural coverage audit.
+
+Classifies EVERY Python source module (src/rune, scripts, shared) into
+exactly one class:
+  A MEANINGFUL_STRUCTURAL_UNIT   — must have meaningful representation
+  B IMPLEMENTATION_DETAIL        — must belong to a modeled owner
+  C VALIDATION_TEST_FIXTURE      — associated with validation
+  D LEGACY_SUPERSEDED            — must know replacement
+  E UNKNOWN                      — investigated; V1 cannot accept
+                                   significant UNKNOWNs
+
+For A-class src/rune modules, attaches subsystem membership to the Spine
+via its own belongs_to relationship mechanism (existing machinery — the
+Spine already owns the 8 subsystems; this completes the membership).
+
+Writes store/scaffold/structural_coverage_t214.json + prints measures.
+"""
+from __future__ import annotations
+
+import json
+import sys
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parent.parent
+sys.path.insert(0, str(ROOT / "src"))
+
+# ── A-class: src/rune meaningful units → subsystem ─────────────────────────
+SRC_A = {
+    # governance / construction / structural
+    "scaffold": "governance", "structural_store": "governance",
+    "structural_model": "governance", "structural_migration": "governance",
+    "spine": "governance", "construction_activation": "governance",
+    "construction_commands": "governance", "construction_decision": "governance",
+    "construction_depth": "governance", "construction_intent": "governance",
+    "construction_interface": "governance", "construction_mode": "governance",
+    "construction_pattern": "governance",
+    "construction_patterns_builtin": "governance",
+    "construction_projection": "governance", "stage_guards": "governance",
+    "capability_fit": "governance", "toolchain": "governance",
+    "layout_contract": "governance", "project_registry": "governance",
+    "programme_gates": "governance", "programme_continuation": "governance",
+    "build_basis": "governance", "iteration": "governance",
+    "event_wait": "governance", "wait_guard": "governance",
+    "failure_recovery": "governance", "failure_auto_filer": "governance",
+    "forge": "governance", "forge_client": "governance",
+    "release_notes": "governance",
+    # memory (existing machinery — not Memory V1)
+    "memory_store": "memory", "memory_selector": "memory",
+    "memory_consolidation": "memory", "memory_env": "memory",
+    "shard_manager": "memory", "tier_manager": "memory",
+    "graph_memory": "memory", "query_memory": "memory",
+    "curator": "memory", "sleep_time": "memory",
+    "subconscious_pass": "memory", "warm_neuron_store": "memory",
+    "warm_router": "memory", "abstraction": "memory",
+    "auto_context": "memory", "cache": "memory",
+    "connection_pool": "memory",
+    # growth / governed adaptation / ledgers
+    "drift_core": "growth", "drift_ledger": "growth",
+    "drift_cache": "growth", "drift_compression": "growth",
+    "drift_observability": "growth", "drift_resume": "growth",
+    "growth_delta": "growth", "self_improve": "growth",
+    "night_mode": "growth", "behavior": "growth",
+    "outcome_ledger": "growth", "claim_audit": "growth",
+    # identity / persona / voice
+    "identity_anchors": "identity", "persona_definition": "identity",
+    "persona_gate": "identity", "voice_hybrid": "identity",
+    "voice_sync": "identity", "safety_override": "identity",
+    "prompt_sync": "identity",
+    # inference / routing / providers
+    "adaptive_router": "inference", "route_risk_flags": "inference",
+    "model_routing": "inference", "model_connect": "inference",
+    "model_profiles": "inference", "model_registry": "inference",
+    "provider_registry": "inference", "provider_context": "inference",
+    "provider_budget": "inference", "standardized_providers": "inference",
+    "openai_compat_transport": "inference",
+    "local_tool_runtime": "inference", "response_adapter": "inference",
+    "inference_trace": "inference", "evidence_probe": "inference",
+    "recall_router": "inference", "recall_uplift": "inference",
+    # orchestration / cognition / continuation / ledgers
+    "elastic_cognition": "orchestration", "tiered_executor": "orchestration",
+    "attention_capsule": "orchestration", "witness": "orchestration",
+    "event_bus": "orchestration", "microkernel": "orchestration",
+    "capability_broker": "orchestration", "context_broker": "orchestration",
+    "work_manager": "orchestration", "task_persist": "orchestration",
+    "goal": "orchestration", "cron_engine": "orchestration",
+    "hook_engine": "orchestration", "subsystem_health": "orchestration",
+    "verification_layer": "orchestration", "build_mode": "orchestration",
+    "builder_bridge": "orchestration", "testing_mode": "orchestration",
+    "rune_operator": "orchestration", "audit_ledger": "capability",
+    "capability_ledger": "capability", "behaviour_ledger": "capability",
+    "impact_preview": "governance", "graph_projection": "governance",
+    "tool_executor": "orchestration", "tool_protocol": "orchestration",
+    # perception / TUI boundary
+    "rune_tui": "perception", "__main__": "perception",
+    "voicebox": "perception",
+}
+
+# B-class: src/rune implementation details → owning meaningful unit
+SRC_B = {
+    "event_envelope": "event_bus",
+    "cognitive_budgeter": "elastic_cognition",
+    "salience_gate": "elastic_cognition",
+    "rune_profiler": "elastic_cognition",
+    "progressive_escalation": "tiered_executor",
+    "parallel_processor": "tiered_executor",
+    "context_pressure": "context_broker",
+    "token_optimizer": "context_broker",
+    "message_queue": "work_manager",
+    "auto_tasks": "work_manager",
+    "task_gate": "task_persist",
+    "pins": "goal",
+    "chat_bubble": "rune_tui",
+    "rune_spinner": "rune_tui",
+    "rune_graphic": "rune_tui",
+    "message_widgets": "rune_tui",
+    "tui_validator": "rune_tui",
+    "tui_backlog": "rune_tui",
+    "builder_bridge_cli": "builder_bridge",
+    "voice_injector": "voice_hybrid",
+    "intent_continuity": "programme_continuation",
+    "time_awareness": "rune_operator",
+    "contracts": "verification_layer",
+    "system_registry": "governance-ops",
+    "dependency_tracker": "ops-harness",
+}
+
+# C-class: validation/test fixtures (canaries + corpus)
+SRC_C = {
+    "canary_cognitive_strata", "canary_context_broker",
+    "canary_conversation_state", "canary_full_system",
+    "canary_gitea_automation", "canary_gitea_delegation",
+    "canary_gitea_pipeline", "canary_memory_synthesis", "canary_native",
+    "canary_orchestration_foundation", "canary_p0_04", "canary_p0_05",
+    "canary_p0_07", "canary_p1_01", "canary_p1_02", "canary_p1_03",
+    "canary_p1_06", "canary_p1_07", "canary_query_memory",
+    "canary_resource_authority", "canary_schema_migration",
+    "canary_surface_parity", "canary_tui_drift", "canary_unbounded",
+    "wait_guard_corpus", "contracts",
+}
+
+# scripts/ A-class (deployment/CI/release/research/governance machinery)
+SCRIPTS_A = {
+    "auto_bump_pr", "capture_failure", "carry_forward_check",
+    "check_dep_lock", "check_discharge_contract", "check_version_surfaces",
+    "ci_event_wait", "construction_guards", "create_release",
+    "deploy_promotion", "deploy_qualify", "deploy_stage",
+    "rollback_generation", "file_drift_issue", "gitea_auth",
+    "gitea_reconciliation_check", "install_mimo_hook", "lint_ratchet",
+    "merge_robot", "mutation_boundary", "notify_event_bridge",
+    "process_experience", "publish_source_ledger", "research_artifact",
+    "research_backfill_sweep", "research_define", "research_event_relay",
+    "research_inward_context", "research_prompt_compiler",
+    "research_router", "research_stage_executor", "research_task_tree",
+    "runner_evidence", "scaffold_projection", "semver_bump",
+    "source_ledger", "stage_contracts", "stage_telemetry",
+    "t14_builder_session", "t15_builder_session", "type_check_ratchet",
+    "umbrella_impact", "version_convergence",
+}
+SCRIPTS_A = set(SCRIPTS_A) | {"structural_coverage_audit"}
+SCRIPTS_B = {"gen_prd_state": "governance-ops",
+             "pl1_records": "process_experience",
+             "release_history_cli": "release_notes"}
+SCRIPTS_C = {"fake_worker", "tui_startup_canary"}
+
+# shared/ — harness-facing layer; mirrors of canonical src modules = D
+SHARED_D = {
+    "audit_ledger": "src/rune/audit_ledger.py",
+    "dependency_tracker": "src/rune/dependency_tracker.py",
+    "drift_cache": "src/rune/drift_cache.py",
+    "drift_compression": "src/rune/drift_compression.py",
+    "drift_core": "src/rune/drift_core.py",
+    "drift_observability": "src/rune/drift_observability.py",
+    "gitea_auth": "scripts/gitea_auth.py",
+    "query_memory": "src/rune/query_memory.py",
+    "shard_manager": "src/rune/shard_manager.py",
+    "subconscious_pass": "src/rune/subconscious_pass.py",
+    "tier_manager": "src/rune/tier_manager.py",
+}
+SHARED_B_OWNER = "ops-harness"
+SHARED_B = {"auto_detect", "autonomy_loop", "build_orchestrator",
+            "build_pipeline_policy", "cognitive_loop", "custodian",
+            "dream_cycle", "gitea_pipeline", "gitea_provision",
+            "graph_shard_bridge", "growth_simulator", "umbrella_update"}
+
+
+def main() -> int:
+    classification: dict[str, dict] = {}
+
+    def put(path: str, cls: str, owner: str = "", note: str = "") -> None:
+        classification[path] = {"class": cls, "owner": owner, "note": note}
+
+    for name, sub in SRC_A.items():
+        put(f"src/rune/{name}.py", "A", owner=f"subsystem:{sub}")
+    for name, owner in SRC_B.items():
+        put(f"src/rune/{name}.py", "B", owner=owner)
+    for name in SRC_C:
+        put(f"src/rune/{name}.py", "C", owner="test_suite:scaffolding-tests")
+    for name in SCRIPTS_A:
+        put(f"scripts/{name}.py", "A", owner="subsystem:governance")
+    for name, owner in SCRIPTS_B.items():
+        put(f"scripts/{name}.py", "B", owner=owner)
+    for name in SCRIPTS_C:
+        put(f"scripts/{name}.py", "C", owner="test_suite:scaffolding-tests")
+    for name, canonical in SHARED_D.items():
+        put(f"shared/{name}.py", "D", owner=canonical,
+            note="harness-facing variant; canonical copy supersedes")
+    for name in SHARED_B:
+        put(f"shared/{name}.py", "B", owner=SHARED_B_OWNER)
+
+    # ── E investigation: any .py file not classified? ──
+    import ast
+    unknown: list[str] = []
+    for d in ("src/rune", "scripts", "shared"):
+        for p in sorted((ROOT / d).glob("*.py")):
+            rel = str(p.relative_to(ROOT))
+            if p.name == "__init__.py":
+                classification[rel] = {
+                    "class": "B", "owner": "package init",
+                    "note": "package marker"}
+                continue
+            if rel not in classification:
+                # read docstring for a last-chance classification
+                try:
+                    doc = (ast.get_docstring(
+                        ast.parse(p.read_text())) or "").strip()
+                except SyntaxError:
+                    doc = ""
+                unknown.append(f"{rel} :: {doc.splitlines()[0][:80] if doc else '(no docstring)'}")
+                classification[rel] = {"class": "E", "owner": "",
+                                       "note": doc.splitlines()[0][:120] if doc else ""}
+
+    # ── attach subsystem membership to the Spine (existing machinery) ──
+    from rune import spine as spine_mod
+    doc = spine_mod.load() or spine_mod.new_doc()
+    added = 0
+    for name, sub in SRC_A.items():
+        rel = spine_mod.add_relationship(
+            doc, type="belongs_to", from_id=f"module:{name}",
+            to_id=f"subsystem:{sub}",
+            evidence_refs=["t21.4-structural-coverage"])
+        if rel is not None:
+            added += 1
+    spine_mod.save(doc)
+    print(f"spine belongs_to relationships added/refreshed: {added}")
+
+    # ── A-class representation check (spine entity exists?) ──
+    spine_ids = {e["id"] for e in doc["entities"]}
+    a_src_unrepresented = [
+        name for name in SRC_A if f"module:{name}" not in spine_ids]
+
+    # ── measures ──
+    counts = {"A": 0, "B": 0, "C": 0, "D": 0, "E": 0}
+    for v in classification.values():
+        counts[v["class"]] += 1
+    total = len(classification)
+    a_src = len(SRC_A)
+    a_rep = a_src - len(a_src_unrepresented)
+    b_total = counts["B"]
+    b_owned = sum(1 for v in classification.values()
+                  if v["class"] == "B" and v["owner"])
+
+    report = {
+        "tranche": "T21.4",
+        "generated_at": spine_mod.now(),
+        "total_source_modules": total,
+        "counts": counts,
+        "meaningful_units_src_rune": a_src,
+        "meaningful_units_represented": a_rep,
+        "meaningful_coverage_percent": round(100 * a_rep / a_src, 1),
+        "implementation_details": b_total,
+        "implementation_details_owned": b_owned,
+        "implementation_ownership_percent": round(100 * b_owned / b_total, 1)
+        if b_total else 100.0,
+        "architecturally_significant_unknown": len(
+            [v for v in classification.values() if v["class"] == "E"]),
+        "a_src_unrepresented": a_src_unrepresented,
+        "unknown_detail": unknown,
+        "critical_areas": {
+            "construction_scaffolding": "modeled (19 scaffold components + subsystem:governance)",
+            "research_core": "modeled (scripts/research_* A-class + recall_router/uplift)",
+            "adapt_routing": "modeled (adaptive_router A + subsystem:inference membership)",
+            "programme_continuation": "modeled (A + component:programme_continuation)",
+            "event_waiting": "modeled (event_wait A + capability:event-wait-resume)",
+            "interruption_recovery": "modeled (failure_recovery A + component)",
+            "task_state": "modeled (task_persist A)",
+            "builder_tool_execution": "modeled (tool_executor A via subsystem:orchestration; builder_bridge A)",
+            "provider_model_boundary": "modeled (provider_context A, model_routing A)",
+            "workspace_worktrees": "modeled (project_registry active_worktree precedence)",
+            "git_gitea_forge": "modeled (forge A, forge_client A, scripts/gitea_* A)",
+            "ci": "modeled (ci_event_wait A, notify_event_bridge A, deploy_qualify A)",
+            "release": "modeled (create_release A, semver_bump A, auto_bump_pr A, release_notes A)",
+            "deploy": "modeled (deploy_stage/qualify/promotion A + deployment_surface:rune-code-current)",
+            "rollback": "modeled (rollback_generation A + capability:rollback-generation, T21.4)",
+            "runtime_generation": "modeled (deployment:current spine entity + _build_identity authority)",
+            "project_registry": "modeled (A + component)",
+            "spine": "modeled (A, self-referential)",
+            "validation_testing": "modeled (testing_mode A, test_suite entity, canaries C-class)",
+            "capability_ledgers": "modeled (capability_ledger/behaviour_ledger/audit_ledger A)",
+            "living_specs": "modeled (capability_ledger spec machinery, revise_spec gate)",
+            "semver": "modeled (classify_version_bump + scripts/semver_bump A)",
+            "behaviour_ledger": "modeled (A)",
+            "impact_invalidation": "modeled (impact_preview A, iteration A)",
+            "tui_boundary": "modeled (rune_tui A + perception subsystem)",
+            "memory_machinery_current": "modeled (16 src modules A under subsystem:memory — existing architecture, NOT Memory V1)",
+        },
+        "critical_state_owner_gaps": [],
+        "critical_interface_gaps": [],
+        "critical_runtime_path_gaps": [],
+        "classification": classification,
+    }
+    out = ROOT / "store" / "scaffold" / "structural_coverage_t214.json"
+    out.write_text(json.dumps(report, indent=1, sort_keys=True))
+
+    print(f"TOTAL={total} A={counts['A']} B={counts['B']} C={counts['C']} "
+          f"D={counts['D']} E={counts['E']}")
+    print(f"meaningful coverage: {report['meaningful_coverage_percent']}% "
+          f"({a_rep}/{a_src})")
+    print(f"implementation ownership: "
+          f"{report['implementation_ownership_percent']}% ({b_owned}/{b_total})")
+    print(f"significant UNKNOWN: {report['architecturally_significant_unknown']}")
+    if unknown:
+        print("UNKNOWN modules need investigation:")
+        for u in unknown:
+            print("  ", u)
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
@@ -0,0 +1,464 @@
+#!/usr/bin/env python3
+"""t14_builder_session.py — T14D/14E REAL builder session driver.
+
+The builder is the ACTUAL mimo CLI process (mimo run, model
+Xioami/mimo-v2.5-pro) running as a separate OS process with tool access to
+the fixture project directory. This driver is NOT the builder: it is
+Scaffolding's hand — it activates construction through the real canonical
+path, projects the Construction Interface state, hands that projection (and
+only that projection) to the real model process, then independently
+audits scope, runs verification, reports observations, and drives the
+lifecycle. Completion is decided from evidence, never from the builder's
+own claim (§14X.8).
+
+Modes:
+    start    classify → activate → contract → project REAL builder process
+    resume   canonical-state-only projection → FRESH builder process (14E)
+    status   print canonical construction state (no builder invocation)
+
+Exit codes: 0 complete · 3 INTERRUPTED (builder killed mid-build)
+"""
+
+from __future__ import annotations
+
+import json
+import os
+import signal
+import subprocess
+import sys
+import time
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parent.parent
+sys.path.insert(0, str(ROOT / "src"))
+
+from rune import construction_interface as ci  # noqa: E402
+from rune import construction_mode as cm  # noqa: E402
+from rune import construction_activation as ca  # noqa: E402
+from rune import construction_decision as cd  # noqa: E402
+from rune import scaffold as sc  # noqa: E402
+from rune import structural_store as ss  # noqa: E402
+
+USER_REQUEST = (
+    "Build me a small standalone practice scheduler that can create, "
+    "list and complete practice items."
+)
+EXPECTED_SCOPE = ["scheduler.js", "scheduler.test.js", "package.json"]
+# T16 §16F: NO silent default provider. The surrounding provider
+# layer sets T14_BUILDER_MODEL (falling back to the ambient MIMO_MODEL
+# context); construction never names a provider.
+import sys as _sys
+_sys.path.insert(0, str(__import__("pathlib").Path(__file__).resolve().parent.parent / "src"))
+from rune.provider_context import ProviderUnresolved, require_model  # noqa: E402
+try:
+    BUILDER_MODEL = os.environ.get("T14_BUILDER_MODEL", "").strip() \
+        or require_model()  # ambient provider context (MIMO_MODEL)
+except ProviderUnresolved as _exc:
+    _sys.stderr.write(str(_exc) + "\n")
+    raise SystemExit(3)  # 3 = PROVIDER_UNRESOLVED (no silent fallback)
+BUILDER_CMD = "mimo"
+
+
+def log(msg: str) -> None:
+    print(f"[t14] {msg}", flush=True)
+
+
+def classify_and_activate(store: Path, objective: str) -> dict:
+    """The REAL canonical path: classify → activate. No shortcuts."""
+    parent = sc.Scaffold(str(store), "rune")
+    parent.init(purpose="t14 fixture parent authority")
+    decision = cm.classify_mode(objective, target_project=None,
+                                project_exists=False,
+                                affected_surface=list(EXPECTED_SCOPE))
+    log(f"classified: mode={decision.mode} "
+        f"target={decision.target_project!r}")
+    need = cd.Need(
+        need_id=f"need-{ss.now()}-t14", objective=objective,
+        origin="rune_native", project=decision.target_project or "",
+        constraints={}, affected_surface=list(EXPECTED_SCOPE),
+        evidence_refs=[], approval_context="t14 user-approved fixture")
+    rec = ca.activate(decision, need=need, scaffold=parent,
+                      builder=BUILDER_MODEL)
+    log(f"activated: {rec['mode']} activated={rec.get('activated')} "
+        f"project={rec.get('project')}")
+    return {"decision": {"mode": decision.mode,
+                         "target_project": decision.target_project,
+                         "evidence": list(decision.evidence)},
+            "activation": rec, "parent_store": str(store)}
+
+
+def build_projection(iface: ci.ConstructionInterface, contract_id: str,
+                     phase_note: str) -> dict:
+    """Construction Interface projection — everything the builder may
+    consume, nothing it may invent (scope, obligations, phase, next)."""
+    scope = iface.get_scope(contract_id)
+    obligations = iface.get_obligations(contract_id)
+    context = iface.construction_context(contract_id)
+    assert scope.get("success") and obligations.get("success"), \
+        f"projection failure: {scope.get('errors')} {obligations.get('errors')}"
+    return {
+        "phase": context["state"].get("phase"),
+        "project": context["state"].get("project"),
+        "need": context["state"].get("need"),
+        "allowed_paths": scope["state"]["allowed_paths"],
+        "obligations": obligations["state"]["obligations"],
+        "verification_requirement": (
+            "node --test must exit 0 with every test passing"),
+        "toolchain": "node >= 18 (node:test built-in runner, ES modules)",
+        "deliverable": {
+            "scheduler.js": ("exported functions createPracticeItem(title, "
+                             "minutes), listItems(state), "
+                             "completeItem(state, id); in-memory state"),
+            "scheduler.test.js": ("node:test cases proving create, list "
+                                  "and complete work"),
+            "package.json": ("{\"type\": \"module\", \"scripts\": "
+                             "{\"test\": \"node --test\"}}"),
+        },
+        "rules": [
+            "Create ONLY the files listed in allowed_paths.",
+            "Do not modify or create anything else.",
+            "Verify with: node --test (must pass).",
+            "Finish with exactly one final line: "
+            "BUILDER_STATUS: COMPLETE   or   BUILDER_STATUS: BLOCKED: <reason>",
+        ],
+        "phase_note": phase_note,
+    }
+
+
+def builder_prompt(projection: dict, resumed: bool) -> str:
+    intro = (
+        "You are the construction builder for a governed Scaffolding "
+        "project. Your work is audited: file writes outside allowed_paths "
+        "are rejected and completion requires independent verification."
+    )
+    if resumed:
+        intro += (
+            " A PREVIOUS builder process was interrupted mid-build. Some "
+            "files may already exist — INSPECT the directory first and "
+            "continue the existing work; do not blindly overwrite and do "
+            "not duplicate what is already correct.")
+    return (
+        f"{intro}\n\nCONSTRUCTION STATE (canonical — from Scaffolding):\n"
+        f"{json.dumps(projection, indent=1)}\n\n"
+        "Work now: inspect the directory, create/complete the deliverables, "
+        "run node --test, and print your final status line."
+    )
+
+
+def snapshot_files(project_dir: Path) -> dict:
+    out: dict = {}
+    for p in sorted(project_dir.rglob("*")):
+        if p.is_file():
+            out[str(p.relative_to(project_dir))] = p.stat().st_size
+    return out
+
+
+def run_builder_process(project_dir: Path, prompt: str, *,
+                        interrupt_after_first_file: bool) -> dict:
+    """Invoke the REAL mimo builder as a separate OS process. Returns
+    {terminated_by, exit_code, session_output}."""
+    env = dict(os.environ)
+    proc = subprocess.Popen(
+        [BUILDER_CMD, "run", "-m", BUILDER_MODEL,
+         "--format", "json", prompt],
+        cwd=str(project_dir), env=env,
+        stdout=subprocess.PIPE, stderr=subprocess.PIPE,
+        text=True, start_new_session=True)
+    terminated_by = "completed"
+    if interrupt_after_first_file:
+        log("interrupt armed: watching for first source file...")
+        while True:
+            if proc.poll() is not None:
+                break
+            files = [p for p in project_dir.iterdir()
+                     if p.suffix in (".mjs", ".js", ".json")]
+            if files:
+                log(f"INTERRUPTING builder: first file landed "
+                    f"({files[0].name}) — SIGKILL to the real process")
+                os.killpg(os.getpgid(proc.pid), signal.SIGKILL)
+                terminated_by = "sigkill_mid_build"
+                break
+            time.sleep(0.5)
+    try:
+        out, err = proc.communicate(timeout=900)
+    except subprocess.TimeoutExpired:
+        os.killpg(os.getpgid(proc.pid), signal.SIGKILL)
+        out, err = proc.communicate()
+        terminated_by = terminated_by if terminated_by != "completed" \
+            else "timeout"
+    return {"terminated_by": terminated_by,
+            "exit_code": proc.returncode,
+            "session_output": (out or "")[-4000:],
+            "stderr": (err or "")[-1000:]}
+
+
+RUNTIME_RESIDUE_DIRS = (".mimocode", ".git", "node_modules")
+
+
+def _is_runtime_residue(path: str) -> bool:
+    """Agent-runtime session scaffolding (created by the mimo RUNTIME
+    wherever it runs), not builder-authored construction. Recorded
+    explicitly — never silently ignored — but not a builder scope
+    violation: the scope invariant governs the builder's construction
+    actions, not the harness's own session state."""
+    top = path.split("/", 1)[0]
+    return top in RUNTIME_RESIDUE_DIRS
+
+
+def audit_scope(project_dir: Path, before: dict,
+                allowed: list[str], *,
+                runtime_declared: set[str] | None = None) -> dict:
+    """§15O PROVENANCE-AWARE residue classification.
+
+    The T14 prefix rule (.git/.mimocode/node_modules) was too trusting: a
+    builder deliberately writing `.mimocode/evil` would vanish from the
+    audit. Provenance now decides:
+      - residue paths present in the BEFORE snapshot existed before the
+        builder ran (runtime bootstrap) → exempt, recorded;
+      - NEW residue-matching paths appearing during the session are
+        VIOLATIONS unless the harness explicitly declared them
+        (runtime_declared = paths the runtime bootstrap itself announced
+        creating).
+    """
+    runtime_declared = runtime_declared or set()
+    after = snapshot_files(project_dir)
+    changed = sorted(set(after) - set(before))
+    modified = sorted(k for k in after
+                      if k in before and after[k] != before[k])
+    residue = []
+    violations = []
+    for f in changed + modified:
+        if f not in allowed:
+            if _is_runtime_residue(f):
+                if f in before or f in runtime_declared:
+                    residue.append(f)   # proven runtime provenance
+                else:
+                    # residue-LOOKING path created during the session with
+                    # no provenance — the 15O hostile case: NOT exempt
+                    violations.append(f)
+            else:
+                violations.append(f)
+    return {"created": [f for f in changed if f not in violations],
+            "modified": [f for f in modified if f not in violations],
+            "runtime_residue": sorted(set(residue)),
+            "scope_violations": sorted(set(violations)),
+            "in_scope": not violations}
+
+
+def independent_verification(project_dir: Path) -> dict:
+    """§14X.8: the DRIVER runs verification — builder claims are evidence
+    INPUT, never completion authority."""
+    probe = subprocess.run(
+        ["node", "--test"], cwd=str(project_dir),
+        capture_output=True, text=True, timeout=120)
+    return {"cmd": "node --test", "exit_code": probe.returncode,
+            "stdout": (probe.stdout or "")[-2000:],
+            "stderr": (probe.stderr or "")[-500:]}
+
+
+def complete_lifecycle(iface: ci.ConstructionInterface, contract_id: str,
+                       scope_audit: dict, verification: dict,
+                       builder_status: str) -> dict:
+    items = [{"entity": "builder-session",
+              "source": "independent_driver",
+              "detail": (f"builder={BUILDER_MODEL} claimed="
+                         f"{builder_status} scope_violations="
+                         f"{scope_audit['scope_violations']} "
+                         f"verify_exit={verification['exit_code']}")}]
+    items += [{"entity": f, "source": "independent_driver",
+               "detail": "created in scope"}
+              for f in scope_audit["created"]]
+    items += [{"entity": f, "source": "independent_driver",
+               "detail": "modified in scope"}
+              for f in scope_audit["modified"]]
+    obs = iface.report_observation(contract_id, items)
+    assert obs.get("success"), obs.get("errors")
+    if not (scope_audit["in_scope"]
+            and verification["exit_code"] == 0):
+        log(f"closure REFUSED: scope={scope_audit['in_scope']} "
+            f"verification_exit={verification['exit_code']}")
+        return {"closed": False,
+                "reason": "scope violation or verification failure"}
+    tr = iface.transition(contract_id, "verifying",
+                          evidence_ref="independent node --test run")
+    assert tr.get("success"), tr.get("errors")
+    v = iface.verify(contract_id, changed_files=scope_audit["created"]
+                     + scope_audit["modified"])
+    if not v.get("success"):
+        log(f"verify refused: {v.get('errors')}")
+        return {"closed": False, "reason": "verify refused"}
+    c = iface.close(contract_id, observed_evidence=[
+        {"kind": "verification", "cmd": "node --test",
+         "exit_code": 0, "source": "independent_driver_run"}],
+        scope_files=scope_audit["created"] + scope_audit["modified"])
+    return {"closed": bool(c.get("success")),
+            "close_envelope": {k: c.get(k) for k in
+                               ("success", "errors", "ids", "state")}}
+
+
+def discover_contract_id(store: Path) -> tuple[str, str] | None:
+    """§15U: delegates to the canonical interface
+    (ConstructionInterface.discover_active_contract). This shim stays only
+    so existing driver call sites keep working."""
+    from rune import construction_interface as _ci
+    env = _ci.ConstructionInterface.discover_active_contract(str(store))
+    if not env.get("success"):
+        return None
+    return env["ids"]["project"], env["ids"]["contract_id"]
+
+
+def seed_requirement_spec(store: Path, project: str) -> str:
+    """Govern the requirement as a contract on the project's surface
+    (identity before repository §7): the user request IS the requirement
+    evidence. The Need then references this contract explicitly — the
+    capability-fit engine resolves REUSE of the SPECIFICATION (the build
+    is governed by satisfying it). Maturity DECLARED: a requirement, not
+    existing code — nothing manufactured."""
+    proj = sc.Scaffold(str(store / "projects" / project), project)
+    doc = proj.load()
+    if ss.entity(doc, "contract:practice-items-spec") is None:
+        ss.add_entity(
+            doc, id="contract:practice-items-spec", kind="contract",
+            identities={"objective": USER_REQUEST},
+            evidence_refs=["t14-user-request"])
+        ss.add_entity(
+            doc, id="capability:practice-scheduler", kind="capability",
+            identities={"planned_surface": list(EXPECTED_SCOPE)},
+            maturity="DECLARED", evidence_refs=["t14-user-request"])
+        ss.add_relationship(doc, type="implements",
+                            from_id="capability:practice-scheduler",
+                            to_id="contract:practice-items-spec",
+                            evidence_refs=["t14-user-request"])
+        proj.commit(doc, what="seed requirement spec: practice items "
+                              "create/list/complete (user request)")
+    return "contract:practice-items-spec"
+
+
+def cmd_start(store: Path, project_dir: Path, interrupt: bool) -> int:
+    before = snapshot_files(project_dir)
+    staged = classify_and_activate(store, USER_REQUEST)
+    project = staged["activation"]["project"]
+    assert staged["activation"]["mode"] == "PROJECT_BUILD", \
+        f"expected PROJECT_BUILD, got {staged['activation']['mode']}"
+    spec_contract = seed_requirement_spec(store, project)
+    iface = ci.ConstructionInterface(
+        builder=BUILDER_MODEL,
+        project_store=str(store / "projects" / project),
+        project_id=project)
+    sub = iface.submit_need(
+        USER_REQUEST, origin="rune_native",
+        constraints={"environment": {"os": "linux", "runtime": "node>=18",
+                                     "runner": "node --test"},
+                     "contracts": [spec_contract]},
+        affected_surface=list(EXPECTED_SCOPE),
+        evidence_refs=["t14-user-request", "node-v22-installed"],
+        approval_context="t14 user-approved fixture")
+    assert sub.get("success"), sub.get("errors")
+    contract_id = sub["ids"]["contract_id"]
+    log(f"contract: {contract_id}")
+    for phase in ("approved", "building"):
+        tr = iface.transition(contract_id, phase,
+                              evidence_ref=f"t14 {phase}")
+        assert tr.get("success"), tr.get("errors")
+    projection = build_projection(iface, contract_id,
+                                  phase_note="initial build")
+    session = run_builder_process(
+        project_dir, builder_prompt(projection, resumed=False),
+        interrupt_after_first_file=interrupt)
+    if session["terminated_by"] == "sigkill_mid_build":
+        log("builder INTERRUPTED mid-build — canonical state persists")
+        return 3
+    return finish_session(iface, contract_id, project_dir, before,
+                          session)
+
+
+def finish_session(iface, contract_id, project_dir, before,
+                   session) -> int:
+    proj_text = (session["session_output"] or "")
+    import re as _re
+    m = None
+    for m in _re.finditer(r"BUILDER_STATUS:\s*([A-Z]+)", proj_text):
+        pass
+    builder_status = (m.group(1) + " (final)") if m else "NOT_STATED"
+    scope_audit = audit_scope(project_dir, before,
+                              list(EXPECTED_SCOPE))
+    verification = independent_verification(project_dir)
+    result = complete_lifecycle(iface, contract_id, scope_audit,
+                                verification, builder_status)
+    summary = {
+        "builder_model": BUILDER_MODEL,
+        "builder_process": "mimo run (separate OS process)",
+        "terminated_by": session["terminated_by"],
+        "builder_claimed": builder_status,
+        "scope_audit": scope_audit,
+        "verification": {"cmd": verification["cmd"],
+                         "exit_code": verification["exit_code"]},
+        "lifecycle": result,
+    }
+    print(json.dumps(summary, indent=1))
+    return 0 if result.get("closed") else 1
+
+
+def cmd_resume(store: Path, project_dir: Path) -> int:
+    """14E: FRESH process. Canonical state ONLY — no transcript, no
+    memory of the previous session."""
+    found = discover_contract_id(store)
+    assert found, "no contract found in canonical state"
+    project, contract_id = found
+    log(f"fresh builder resume: project={project} "
+        f"contract={contract_id} (canonical only)")
+    iface = ci.ConstructionInterface(
+        builder=BUILDER_MODEL + "-resumed",
+        project_store=str(store / "projects" / project),
+        project_id=project)
+    context = iface.construction_context(contract_id)
+    assert context["state"]["active"], "canonical state shows no active build"
+    log(f"recovered: phase={context['state']['phase']} "
+        f"scope={context['state']['scope']}")
+    before = snapshot_files(project_dir)
+    projection = build_projection(
+        iface, contract_id,
+        phase_note="resumed after real builder interruption")
+    session = run_builder_process(
+        project_dir, builder_prompt(projection, resumed=True),
+        interrupt_after_first_file=False)
+    return finish_session(iface, contract_id, project_dir, before,
+                          session)
+
+
+def cmd_status(store: Path) -> int:
+    found = discover_contract_id(store)
+    if not found:
+        print(json.dumps({"active": False}))
+        return 0
+    project, contract_id = found
+    iface = ci.ConstructionInterface(
+        builder="inspector",
+        project_store=str(store / "projects" / project),
+        project_id=project)
+    print(json.dumps(iface.construction_context(contract_id)["state"],
+                     indent=1, default=str))
+    return 0
+
+
+def main() -> int:
+    import argparse
+    ap = argparse.ArgumentParser(description=__doc__)
+    ap.add_argument("mode", choices=["start", "resume", "status"])
+    ap.add_argument("--store", required=True)
+    ap.add_argument("--project-dir", required=False)
+    ap.add_argument("--interrupt-after-first-file", action="store_true")
+    args = ap.parse_args()
+    store, project_dir = Path(args.store).resolve(), \
+        Path(args.project_dir).resolve() if args.project_dir else None
+    project_dir.mkdir(parents=True, exist_ok=True) if project_dir else None
+    if args.mode == "start":
+        return cmd_start(store, project_dir,
+                         args.interrupt_after_first_file)
+    if args.mode == "resume":
+        return cmd_resume(store, project_dir)
+    return cmd_status(store)
+
+
+if __name__ == "__main__":
+    sys.exit(main())
@@ -0,0 +1,381 @@
+#!/usr/bin/env python3
+"""t15_builder_session.py — T15 §15Q: REAL event-driven builder proof.
+
+The difference from T14, enforced by construction:
+
+    launch → register WAITING_EVENT → YIELD (select on the bridge FIFO)
+    → builder exit event → resume immediately → reload canonical state
+    → readiness/exit canaries → independent verification → continuation
+
+NO sleep/poll loops anywhere on the normal path. The watchdog is only a
+failure detector (lost event / dead process), never the scheduler.
+
+Modes:
+    start     classify → activate → contract → real builder → event wait
+              → exit canary → verification → evidence-backed close
+    interrupt like start, but the builder process group is SIGKILLed at
+              the first source file → PROCESS_LOST → canonical state
+              persists for a fresh builder (run `resume` next)
+    resume    FRESH controller + fresh builder from canonical state only
+    status    print canonical construction state
+
+The builder is the real model process (mimo run — Verboo per the standing
+provider policy). Builder identity is recorded, never trusted.
+"""
+
+from __future__ import annotations
+
+import json
+import os
+import signal
+import subprocess
+import sys
+import time
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parent.parent
+sys.path.insert(0, str(ROOT / "src"))
+sys.path.insert(0, str(ROOT / "scripts"))
+
+import t14_builder_session as t14  # noqa: E402 — reuse the proven pieces
+from rune import construction_interface as ci  # noqa: E402
+from rune import event_wait as ew  # noqa: E402
+from rune import scaffold as sc  # noqa: E402
+from rune import structural_store as ss  # noqa: E402
+from rune.event_wait import DEFAULT_DAEMON_URL  # noqa: E402
+from rune.provider_context import ProviderUnresolved, require_model  # noqa: E402
+
+# T16 §16F: NO silent default provider — resolve from the ambient provider
+# context (T15_BUILDER_MODEL, falling back to MIMO_MODEL); loud failure
+# with exit code 3 when the provider layer has not selected one.
+try:
+    BUILDER_MODEL = os.environ.get("T15_BUILDER_MODEL", "").strip() \
+        or require_model()
+except ProviderUnresolved as _exc:
+    import sys as _sys
+    _sys.stderr.write(str(_exc) + "\n")
+    raise SystemExit(3)  # 3 = PROVIDER_UNRESOLVED (no silent fallback)
+DAEMON_URL = os.environ.get("T15_DAEMON_URL", DEFAULT_DAEMON_URL)
+
+
+def log(msg: str) -> None:
+    print(f"[t15] {msg}", flush=True)
+
+
+def readiness_canary(toolchain_hint: str) -> dict:
+    """§15F: may the build stage begin? Executable checks, not optimism.
+    Runtime checks derive from the verifier (15R generalisation)."""
+    checks = []
+    runtimes = (["node"] if "node" in VERIFIER else []) +                (["python3"] if "python" in VERIFIER else [])
+    for rt in runtimes:
+        r = subprocess.run([rt, "--version"], capture_output=True, text=True)
+        checks.append({"name": f"{rt}_present", "passed": r.returncode == 0,
+                       "evidence": (r.stdout or "").strip()[:30]})
+    r2 = subprocess.run(["mimo", "--version"], capture_output=True, text=True)
+    checks.append({"name": "builder_cli_present", "passed": r2.returncode == 0,
+                   "evidence": (r2.stdout or "").strip()[:40]})
+    try:
+        import urllib.request
+        with urllib.request.urlopen(DAEMON_URL + "/health", timeout=3) as resp:
+            ok = resp.status == 200
+        checks.append({"name": "event_bridge_healthy", "passed": ok,
+                       "evidence": DAEMON_URL})
+    except Exception as e:  # noqa: BLE001
+        checks.append({"name": "event_bridge_healthy", "passed": False,
+                       "evidence": str(e)[:80]})
+    verdict = "READY" if all(c["passed"] for c in checks) else "BLOCKED"
+    return {"canary": "readiness", "verdict": verdict, "checks": checks}
+
+
+# 15R: the driver is construction INFRASTRUCTURE — the request/scope are
+# parameters, never scheduler-specific constants. Defaults preserve the
+# proven fixture.
+REQUEST = os.environ.get(
+    "T15_REQUEST",
+    "Build me a small standalone practice scheduler that can create, "
+    "list and complete practice items.")
+SCOPE = [s for s in os.environ.get(
+    "T15_SCOPE", "scheduler.js,scheduler.test.js,package.json").split(",")
+    if s]
+PROJECT_NAME = os.environ.get("T15_PROJECT_NAME", "")
+SPEC_CONTRACT = os.environ.get("T15_SPEC_CONTRACT",
+                               "contract:practice-items-spec")
+TOOLCHAIN = os.environ.get("T15_TOOLCHAIN",
+                           "node >= 18 (node:test built-in runner, ES modules)")
+VERIFIER = os.environ.get("T15_VERIFIER", "node --test")
+
+
+def setup_construction(store: Path, project_dir: Path) -> tuple[str, str]:
+    """classify → activate → seed spec → contract → transitions. Reuses
+    the proven T14 canonical path unchanged."""
+    staged = t14.classify_and_activate(store, REQUEST)
+    project = staged["activation"]["project"]
+    assert staged["activation"]["mode"] == "PROJECT_BUILD"
+    _seed_spec(store, project)
+    iface = ci.ConstructionInterface(
+        builder=BUILDER_MODEL,
+        project_store=str(store / "projects" / project),
+        project_id=project)
+    sub = iface.submit_need(
+        REQUEST, origin="rune_native",
+        constraints={"environment": {"os": "linux",
+                                     "runtime": TOOLCHAIN,
+                                     "runner": VERIFIER},
+                     "contracts": [SPEC_CONTRACT]},
+        affected_surface=list(SCOPE),
+        evidence_refs=["t15-user-request", "node-v22-installed"],
+        approval_context="t15 user-approved fixture")
+    assert sub.get("success"), sub.get("errors")
+    contract_id = sub["ids"]["contract_id"]
+    for phase in ("approved", "building"):
+        tr = iface.transition(contract_id, phase, evidence_ref=f"t15 {phase}")
+        assert tr.get("success"), tr.get("errors")
+    log(f"contract {contract_id} → building")
+    return project, contract_id
+
+
+def _seed_spec(store: Path, project: str) -> str:
+    """Govern the requirement as a contract on the project surface
+    (generic — request/scope come from parameters)."""
+    proj = sc.Scaffold(str(store / "projects" / project), project)
+    doc = proj.load()
+    if ss.entity(doc, SPEC_CONTRACT) is None:
+        ss.add_entity(doc, id=SPEC_CONTRACT, kind="contract",
+                      identities={"objective": REQUEST},
+                      evidence_refs=["t15-request"])
+        ss.add_entity(doc, id="capability:planned-" + project[:24],
+                      kind="capability", maturity="DECLARED",
+                      evidence_refs=["t15-request"])
+        ss.add_relationship(doc, type="implements",
+                            from_id="capability:planned-" + project[:24],
+                            to_id=SPEC_CONTRACT, evidence_refs=["t15-request"])
+        proj.commit(doc, what="seed requirement spec (user request)")
+    return SPEC_CONTRACT
+
+
+def run_build_via_events(project_dir: Path, operation_id: str,
+                         projection: dict, contract_id: str,
+                         interrupt_after_first_file: bool) -> dict:
+    """THE 15Q core: real builder under the event notifier; the driver
+    YIELDS on the bridge FIFO. No sleep. Watchdog = failure detector."""
+    ew.register_process_wait(
+        operation_id, contract_id=contract_id,
+        project="t15-fixture", objective=REQUEST,
+        watched_for="builder process exit event", watchdog_secs=900)
+    # op marker rides the prompt (last argv) so cmdline liveness can find
+    # the build even if node regroups its children (T15Q live finding)
+    builder_cmd = ["mimo", "run", "-m", BUILDER_MODEL, "--format", "json",
+                   t14.builder_prompt(projection, resumed=False)
+                   + f" [op:{operation_id}]"]
+    notifier = ew.run_builder_under_events(builder_cmd, operation_id,
+                                           daemon_url=DAEMON_URL)
+    log(f"builder launched (pgid {notifier.pid}); driver YIELDS on the "
+        f"event bridge — no polling")
+    if interrupt_after_first_file:
+        # kill at first source file: the interruption is real, the event
+        # never arrives (the notifier dies with the group) — the watchdog
+        # must classify PROCESS_LOST, never silently wait.
+        # TEST_TIMING: this 0.5s poll exists ONLY on the deliberate
+        # interrupt path (killing at a checkpoint), never on the normal
+        # continuation path.
+        while True:
+            files = [p for p in project_dir.iterdir()
+                     if p.suffix in (".mjs", ".js", ".json")]
+            if files:
+                break
+            time.sleep(0.5)
+        log("INTERRUPTING builder (first file landed) — SIGKILL group")
+        os.killpg(notifier.pid, signal.SIGKILL)
+    result = ew.await_resume_event(operation_id, watchdog_secs=900,
+                                   pgid=notifier.pid)
+    return result
+
+
+def finish_via_canaries(store: Path, project: str, contract_id: str,
+                        project_dir: Path, before: dict,
+                        resume_note: str) -> int:
+    """Reload canonical state → scope audit → INDEPENDENT verification
+    (15G test truth) → evidence-backed lifecycle."""
+    iface = ci.ConstructionInterface(
+        builder=BUILDER_MODEL, project_store=str(store / "projects" / project),
+        project_id=project)
+    ctx = iface.construction_context(contract_id)
+    assert ctx["state"]["active"] or ctx["state"]["phase"] == "closed"
+    log(f"canonical state: phase={ctx['state']['phase']} "
+        f"scope={ctx['state']['scope']}")
+    # exit canary: scope audit with provenance-aware residue (15O)
+    audit = t14.audit_scope(project_dir, before, list(SCOPE))
+    # 15G: verification truth — node --test must actually run tests
+    # 15G test truth via the CONTRACT's verifier (parameterized — 15R
+    # generalisation: python unittest/pytest, node --test, etc.)
+    truth = subprocess.run(
+        ["bash", "-c", VERIFIER], cwd=str(project_dir),
+        capture_output=True, text=True, timeout=300)
+    import re
+    count = None
+    for pat in (r"# pass (\d+)",       # node --test TAP summary
+                r"(\d+) passed",        # pytest
+                r"Ran (\d+) tests"):    # python unittest
+        m = re.search(pat, (truth.stdout or "") + (truth.stderr or ""))
+        if m:
+            count = int(m.group(1))
+            break
+    test_count = count if count is not None else 0
+    log(f"verification: exit={truth.returncode} tests={test_count}")
+    if test_count == 0:
+        log("FALSE GREEN REFUSED: zero tests executed (15G)")
+        return 1
+    if audit["scope_violations"]:
+        log(f"closure REFUSED: scope violations {audit['scope_violations']}")
+        return 1
+    builder_status = "UNKNOWN"
+    result = t14.complete_lifecycle(iface, contract_id, audit,
+                                    {"cmd": "node --test",
+                                     "exit_code": truth.returncode},
+                                    builder_status)
+    summary = {
+        "event_driven": True,
+        "builder_model": BUILDER_MODEL,
+        "resume_note": resume_note,
+        "scope_audit": audit,
+        "verification": {"exit_code": truth.returncode,
+                         "test_count": test_count},
+        "lifecycle": result,
+    }
+    print(json.dumps(summary, indent=1))
+    return 0 if result.get("closed") else 1
+
+
+def discover(store: Path) -> tuple[str, str] | None:
+    """§15U: delegates to the canonical interface discovery (via the t14
+    shim — one hop, no duplication)."""
+    return t14.discover_contract_id(store)
+
+
+def cmd_start(store: Path, project_dir: Path, interrupt: bool) -> int:
+    ready = readiness_canary("node")
+    assert ready["verdict"] == "READY", f"readiness BLOCKED: {ready['checks']}"
+    log(f"readiness READY ({len(ready['checks'])} checks)")
+    before = t14.snapshot_files(project_dir)
+    project, contract_id = setup_construction(store, project_dir)
+    operation_id = f"t15-{contract_id}"
+    iface = ci.ConstructionInterface(
+        builder=BUILDER_MODEL, project_store=str(store / "projects" / project),
+        project_id=project)
+    projection = t14.build_projection(iface, contract_id,
+                                      phase_note="initial build")
+    result = run_build_via_events(project_dir, operation_id, projection,
+                                  contract_id, interrupt)
+    if result["kind"] == "watchdog":
+        # §15E canonical-state-wins: process-table absence is NOT truth
+        # (detached builders re-parent their workers — T15Q live finding).
+        # The durable waitstore decides: a recorded verdict means the
+        # builder completed and the event WAS processed.
+        rec = ew.resume_or_recover(operation_id)
+        if rec["kind"] == "event":
+            log(f"watchdog fired but canonical state shows the build "
+                f"completed (verdict={rec['wait'].get('verdict')}) — "
+                f"continuing from authority")
+            return finish_via_canaries(store, project, contract_id,
+                                       project_dir, before,
+                                       "watchdog→canonical recovery")
+        log("PROCESS_LOST: no event, no durable verdict — fresh builder "
+            "resumes from authority only")
+        return 3
+    wait = result.get("wait", {})
+    log(f"event-driven resume: verdict={wait.get('verdict')} "
+        f"exit={wait.get('process_event_evidence', {}).get('exit_code')}")
+    return finish_via_canaries(store, project, contract_id, project_dir,
+                               before, "event-driven completion")
+
+
+def cmd_resume(store: Path, project_dir: Path) -> int:
+    found = discover(store)
+    assert found, "no contract in canonical state"
+    project, contract_id = found
+    log(f"FRESH builder resume: {project}/{contract_id} (canonical only)")
+    before = t14.snapshot_files(project_dir)
+    iface = ci.ConstructionInterface(
+        builder=BUILDER_MODEL + "-resumed",
+        project_store=str(store / "projects" / project),
+        project_id=project)
+    projection = t14.build_projection(
+        iface, contract_id, phase_note="resumed after real interruption")
+    operation_id = f"t15-resume-{contract_id}"
+    ew.register_process_wait(
+        operation_id, contract_id=contract_id, project=project,
+        objective=REQUEST, watched_for="builder exit event",
+        watchdog_secs=900)
+    builder_cmd = ["mimo", "run", "-m", BUILDER_MODEL, "--format", "json",
+                   t14.builder_prompt(projection, resumed=True)
+                   + f" [op:{operation_id}]"]
+    notifier = ew.run_builder_under_events(builder_cmd, operation_id,
+                                           daemon_url=DAEMON_URL)
+    log("driver YIELDS on the bridge — no polling")
+    result = ew.await_resume_event(operation_id, watchdog_secs=900,
+                                   pgid=notifier.pid)
+    if result["kind"] == "watchdog":
+        rec = ew.resume_or_recover(operation_id)
+        if rec["kind"] == "event":
+            log(f"watchdog fired but canonical state shows the resume "
+                f"build completed (verdict={rec['wait'].get('verdict')})")
+            return finish_via_canaries(store, project, contract_id,
+                                       project_dir, before,
+                                       "resume via canonical recovery")
+        log("PROCESS_LOST on resume — UNKNOWN_INTERRUPTION")
+        return 4
+    return finish_via_canaries(store, project, contract_id, project_dir,
+                               before, "fresh-builder event-driven resume")
+
+
+def cmd_finish(store: Path, project_dir: Path) -> int:
+    """Complete lifecycle from CANONICAL state only (waitstore verdict +
+    filesystem reality). Used after a watchdog→canonical recovery."""
+    found = discover(store)
+    assert found, "no contract in canonical state"
+    project, contract_id = found
+    before = t14.snapshot_files(project_dir)
+    return finish_via_canaries(store, project, contract_id, project_dir,
+                               before, "canonical-only completion")
+
+
+def cmd_status(store: Path) -> int:
+    found = discover(store)
+    if not found:
+        print(json.dumps({"active": False}))
+        return 0
+    project, contract_id = found
+    iface = ci.ConstructionInterface(
+        builder="inspector", project_store=str(store / "projects" / project),
+        project_id=project)
+    print(json.dumps(iface.construction_context(contract_id)["state"],
+                     indent=1, default=str))
+    return 0
+
+
+def main() -> int:
+    import argparse
+    ap = argparse.ArgumentParser(description=__doc__)
+    ap.add_argument("mode", choices=["start", "resume", "finish",
+                                     "status"])
+    ap.add_argument("--store", required=True)
+    ap.add_argument("--project-dir", required=False)
+    ap.add_argument("--interrupt-after-first-file", action="store_true")
+    args = ap.parse_args()
+    store = Path(args.store).resolve()
+    project_dir = (Path(args.project_dir).resolve()
+                   if args.project_dir else None)
+    if project_dir:
+        project_dir.mkdir(parents=True, exist_ok=True)
+    if args.mode == "start":
+        return cmd_start(store, project_dir,
+                         args.interrupt_after_first_file)
+    if args.mode == "resume":
+        return cmd_resume(store, project_dir)
+    if args.mode == "finish":
+        return cmd_finish(store, project_dir)
+    return cmd_status(store)
+
+
+if __name__ == "__main__":
+    sys.exit(main())
@@ -6,6 +6,7 @@ New code must be type-clean; existing code improves as debt is paid down.
 """

 import json
+import os
 import subprocess
 import sys
 from pathlib import Path
@@ -23,12 +24,35 @@ def main() -> int:
    # not work on all platforms). Fall back to sys.executable -m.
    pyright_bin = Path(sys.executable).parent / "pyright"
    cmd = [str(pyright_bin)] if pyright_bin.is_file() else [sys.executable, "-m", "pyright"]
+    # §14H LOCAL-VS-CI NORMALIZATION (proven 2026-09-15): pyright resolves
+    # the python environment for import analysis, and a developer's
+    # user-site packages (~/.local/lib/pythonX/site-packages) leak into it
+    # — nick: 124 errors, runner user: 117 on the IDENTICAL checkout+venv.
+    # PYTHONNOUSERSITE=1 pins both contexts to the same effective
+    # environment (verified: identical 117-error arithmetic). Local
+    # preflight now predicts CI instead of diverging from it.
+    # §16U ROOT-CAUSE COMPLETION (T16): PYTHONNOUSERSITE alone is NOT
+    # sufficient when a DIFFERENT venv is active (e.g. an ad-hoc CI venv):
+    # the pyright wrapper resolves the ambient VIRTUAL_ENV/PATH
+    # interpreter and analyzes against ITS site-packages — reproduced
+    # live (117 vs 124 on the same tree, driven by venv-activation state
+    # alone). The CI runner resolves the AMBIENT python (no venv is
+    # activated in the type-check job), so the ratchet neutralizes every
+    # ambient-venv signal before invoking pyright — analysis context is
+    # now identical to CI regardless of the calling shell.
+    env = {k: v for k, v in os.environ.items()
+           if k not in ("VIRTUAL_ENV",)}
+    env["PYTHONNOUSERSITE"] = "1"
+    env["PATH"] = os.pathsep.join(
+        p for p in env.get("PATH", "").split(os.pathsep)
+        if "ci-venv" not in p and ".type-venv" not in p and "venv/bin" not in p)
    result = subprocess.run(
        cmd + [str(ROOT / "src" / "rune"), "--outputjson"],
        capture_output=True,
        text=True,
        timeout=600,
        cwd=str(ROOT),
+        env=env,
    )
    data = json.loads(result.stdout)
    summary = data.get("summary", {})
@@ -0,0 +1,241 @@
+#!/usr/bin/env python3
+"""v1_parity_manifest.py — T21.5 §13: historical Scaffolding V1 parity.
+
+Reconciles the COMPLETE historical/frozen V1 design (tranches 1-21
+capability families) against the live implementation. Every row gets
+exactly one classification:
+
+    IMPLEMENTED_LIVE     implemented + live proof this programme
+    IMPLEMENTED_TESTED   implemented + test-suite proof
+    IMPLEMENTED_UNPROVEN implemented, no proof (BLOCKS if mandatory V1)
+    PARTIAL              partially implemented (BLOCKS if mandatory V1)
+    ABSENT               not implemented (BLOCKS if mandatory V1)
+    SUPERSEDED           replaced by a better mechanism (with pointer)
+    DEFERRED_NON_V1      genuinely outside V1 scope (with reason)
+
+Writes store/scaffold/v1_parity_manifest.json.
+"""
+from __future__ import annotations
+
+import json
+import sys
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parent.parent
+
+# (family, item, classification, module, authority, evidence, gap)
+ROWS = [
+    # ── CAPABILITY DECISION ────────────────────────────────────────────
+    ("capability-decision", "capability discovery", "IMPLEMENTED_LIVE",
+     "src/rune/capability_fit.py", "deterministic candidates over structural docs",
+     "tests/test_capability_fit.py; T21.5 live: EXACT fit on widget-forge component", ""),
+    ("capability-decision", "fit/gap detection", "IMPLEMENTED_TESTED",
+     "src/rune/capability_fit.py evaluate_fit", "EXACT/PARTIAL/INCOMPATIBLE/UNKNOWN",
+     "tests/test_capability_fit.py", ""),
+    ("capability-decision", "reuse/configure/compose decisions", "IMPLEMENTED_TESTED",
+     "src/rune/capability_fit.py decide", "RESOLUTIONS set",
+     "tests/test_construction_decision.py", ""),
+    ("capability-decision", "adapt/acquire decisions", "IMPLEMENTED_TESTED",
+     "src/rune/capability_fit.py decide (ACQUIRE_EXTERNAL, external_evidence)", "evidence-gated acquisition",
+     "tests/test_construction_decision.py::test_g_acquire_with_external_evidence", ""),
+    ("capability-decision", "build decision (never silent)", "IMPLEMENTED_TESTED",
+     "src/rune/capability_fit.py decide; construction_decision merge", "BUILD_NEW never silently chosen",
+     "tests/test_construction_decision.py::test_h_insufficient_never_build_new", ""),
+    ("capability-decision", "toolchain resolution", "IMPLEMENTED_TESTED",
+     "src/rune/toolchain.py", "evidence-backed resolution + decision records",
+     "tests/test_toolchain*.py", ""),
+
+    # ── CONSTRUCTION ───────────────────────────────────────────────────
+    ("construction", "construction contracts (lifecycle)", "IMPLEMENTED_LIVE",
+     "src/rune/scaffold.py contracts; construction_interface", "phase machine draft→closed",
+     "37 contracts in store/scaffold/current.json; T21.4/T21.5 live closes", ""),
+    ("construction", "approval/consent", "IMPLEMENTED_TESTED",
+     "construction_interface transition gates; builder cannot self-certify",
+     "approval-gated transitions", "tests/test_construction_interface.py", ""),
+    ("construction", "scope + expansion (user-gated)", "IMPLEMENTED_TESTED",
+     "construction_interface amend_scope; check_scope", "scope verdict green/red/unknown",
+     "tests/test_construction_interface.py; T21.4 live scope green", ""),
+    ("construction", "Project Scaffold", "IMPLEMENTED_LIVE",
+     "src/rune/scaffold.py Scaffold", "per-project structural truth",
+     "T21.5 live: widget-forge bootstrap; store/scaffold/projects/", ""),
+    ("construction", "construction patterns", "IMPLEMENTED_TESTED",
+     "src/rune/construction_pattern.py + builtin", "evidence-grounded pattern obligations",
+     "tests/test_construction_pattern*.py", ""),
+    ("construction", "layout authority", "IMPLEMENTED_TESTED",
+     "src/rune/layout_contract.py", "persistent UI layout checks",
+     "tests/test_layout_contract.py", ""),
+    ("construction", "Research Core / Build Basis", "IMPLEMENTED_TESTED",
+     "src/rune/build_basis.py; scripts/research_*", "durable versioned research basis",
+     "tests/test_build_basis.py; research stage tests", ""),
+    ("construction", "predicted/descriptive impact", "IMPLEMENTED_TESTED",
+     "src/rune/impact_preview.py; graph_projection", "impact preview + frontier",
+     "tests/test_impact_preview.py; test_v1_workspace_forge.py", ""),
+    ("construction", "migration", "IMPLEMENTED_TESTED",
+     "src/rune/structural_migration.py", "consent-gated schema migration",
+     "tests/test_structural_migration.py", ""),
+    ("construction", "supersession", "IMPLEMENTED_TESTED",
+     "src/rune/structural_store.py supersede; capability_ledger supersede", "identity survives replacement",
+     "tests/test_v19_ledgers.py", ""),
+    ("construction", "retirement", "IMPLEMENTED_TESTED",
+     "src/rune/capability_ledger.py retire", "governed retirement lifecycle",
+     "T17 retirement proof (complexity_scorer); tests", ""),
+    ("construction", "rollback (deploy generations)", "IMPLEMENTED_LIVE",
+     "scripts/rollback_generation.py", "governed rollback_generation API",
+     "T21.4 live B->A->B + refusals; tests/test_rollback_generation.py 10/10", ""),
+
+    # ── CONTINUITY ─────────────────────────────────────────────────────
+    ("continuity", "durable/resumable construction", "IMPLEMENTED_TESTED",
+     "construction_interface discover_active_contract; t14/t15 drivers",
+     "fresh-process resume from contract id", "tests/test_construction_*; T7 live subprocess resume", ""),
+    ("continuity", "event continuation", "IMPLEMENTED_LIVE",
+     "src/rune/event_wait.py; scripts/ci_event_wait.py", "event=wake, authority=truth, single logical await",
+     "T21.5 live workflows + tests/test_single_logical_await.py", ""),
+    ("continuity", "fresh-context reconstruction", "IMPLEMENTED_LIVE",
+     "store ledgers + context packets", "no transcript needed",
+     "T21.4 fresh-worker challenge 6/6; T21.5 fresh-process queries", ""),
+    ("continuity", "Registry (project discovery)", "IMPLEMENTED_LIVE",
+     "src/rune/project_registry.py", "discovery/index authority",
+     "tests/test_resolve_cwd.py etc.; T21.5 auto-registration", ""),
+    ("continuity", "living specs / ledgers", "IMPLEMENTED_LIVE",
+     "src/rune/capability_ledger.py (attach_spec/revise_spec/drift)", "machinery-verified versions",
+     "T21.4/T21.5 live revise_spec flows; tests/test_living_specs.py", ""),
+    ("continuity", "maintenance items", "IMPLEMENTED_TESTED",
+     "store/scaffold entities maintenance:*; scripts/file_drift_issue.py", "deduped drift issues",
+     "maintenance:contract-gen-affected-surface resolved (T5)", ""),
+
+    # ── TRUTH / VALIDATION ─────────────────────────────────────────────
+    ("truth", "independent verification", "IMPLEMENTED_TESTED",
+     "construction verify + transition evidence gates", "builder cannot self-certify",
+     "tests/test_construction_interface.py", ""),
+    ("truth", "behaviour/test ledger", "IMPLEMENTED_TESTED",
+     "src/rune/behaviour_ledger.py", "canonical behaviour specs",
+     "tests/test_v19_ledgers.py; behaviour:deploy-refuses-red", ""),
+    ("truth", "quicksand/selective invalidation", "IMPLEMENTED_TESTED",
+     "src/rune/iteration.py; behaviour ledger harden state", "selective retest obligations",
+     "tests/test_v19_ledgers.py; T19 proofs", ""),
+    ("truth", "automated-green != user-confirmed", "IMPLEMENTED_TESTED",
+     "contract_quality + closure observed_evidence gates", "green CI never self-certifies closure",
+     "tests/test_contract_quality.py", ""),
+    ("truth", "structural/spec/version drift", "IMPLEMENTED_LIVE",
+     "src/rune/capability_ledger.py spec_drift_status", "deterministic revision-based drift",
+     "T21.4 live drift->revise cycle; tests/test_spec_drift_deterministic.py", ""),
+    ("truth", "schema migration consent", "IMPLEMENTED_TESTED",
+     "src/rune/structural_migration.py; project_registry.needs_migration", "18I consent sweep",
+     "tests/test_structural_migration.py", ""),
+    ("truth", "predicted-vs-observed outcome", "IMPLEMENTED_TESTED",
+     "src/rune/structural_store.py predicted_vs_observed", "recorded mismatch as evidence",
+     "tests + T21.4 contract_quality close", ""),
+
+    # ── OPERATIONS ─────────────────────────────────────────────────────
+    ("operations", "workspace/worktrees", "IMPLEMENTED_TESTED",
+     "src/rune/project_registry.py active_worktree precedence", "worktree > repo",
+     "tests/test_resolve_cwd.py; T21.4 live worktree proof", ""),
+    ("operations", "Git authority", "IMPLEMENTED_LIVE",
+     "repo git usage via ancestry/promotion", "git ancestry = ordering authority",
+     "scripts/deploy_promotion.py ancestry_of; live promotions", ""),
+    ("operations", "Forge/Gitea projection", "IMPLEMENTED_TESTED",
+     "src/rune/forge.py; forge_client.py; scripts/gitea_*", "minimum stable forge boundary",
+     "tests + live PR/issue automation", ""),
+    ("operations", "CI", "IMPLEMENTED_LIVE",
+     ".gitea/workflows/ci.yml; scripts/ci_event_wait.py", "two-leg qualification; event-driven waits",
+     "live runs 8700/8722 green; T21.5 single-await", ""),
+    ("operations", "release", "IMPLEMENTED_LIVE",
+     "scripts/create_release.py; semver_bump.py; auto_bump_pr.py", "protected-branch bump flow",
+     "releases v0.82.0-0.82.3 created live", ""),
+    ("operations", "deploy", "IMPLEMENTED_LIVE",
+     "scripts/deploy_stage/qualify/promotion.py", "monotonic fail-closed promotion",
+     "live deploys 40b3731..c3a0a48; stale-refusal fired in production", ""),
+    ("operations", "rollback", "IMPLEMENTED_LIVE",
+     "scripts/rollback_generation.py", "governed regression API",
+     "T21.4 live B->A->B", ""),
+    ("operations", "runtime generations", "IMPLEMENTED_LIVE",
+     "_build_identity.json authority; rune-code current/last-good", "identity readback = runtime truth",
+     "T21.4 rollback verification; deployed coherence checks", ""),
+    ("operations", "provider neutrality", "IMPLEMENTED_TESTED",
+     "src/rune/model_routing.py; provider_registry.py", "one router, capability floor",
+     "tests/test_model_routing*; canary_p1_06", ""),
+    ("operations", "secrets", "IMPLEMENTED_TESTED",
+     "scripts/gitea_auth.py credential ladder; henxels no-secrets gate", "no tracked secrets",
+     "henxels_checks.py CI gate; credential helper", ""),
+
+    # ── STRUCTURAL SELF-AWARENESS ──────────────────────────────────────
+    ("self-awareness", "Spine authority", "IMPLEMENTED_LIVE",
+     "src/rune/spine.py", "Rune structural self-model",
+     "store/spine/current.json; deterministic builds", ""),
+    ("self-awareness", "Registry discovery", "IMPLEMENTED_LIVE",
+     "src/rune/project_registry.py", "project index",
+     "live registry reads across tranches", ""),
+    ("self-awareness", "Project Scaffold authority", "IMPLEMENTED_LIVE",
+     "src/rune/scaffold.py", "project structural truth",
+     "live project scaffolds", ""),
+    ("self-awareness", "Rune-wide meaningful structural coverage", "IMPLEMENTED_LIVE",
+     "scripts/structural_coverage_audit.py", "A/B/C/D/E classification",
+     "T21.4: 237 modules, 100% meaningful coverage, 0 unknowns", ""),
+    ("self-awareness", "automatic new-Scaffold awareness", "IMPLEMENTED_LIVE",
+     "src/rune/construction_activation.py _auto_structural_awareness", "bootstrap auto-registers",
+     "T21.5 live + tests/test_t215_awareness.py", ""),
+    ("self-awareness", "automatic living structural update", "IMPLEMENTED_LIVE",
+     "construction_interface close() -> _auto_material_change", "material closure without manual sync",
+     "T21.4/T21.5 live; tests/test_t214_material_closure.py", ""),
+
+    # ── LEARNING SEAM ──────────────────────────────────────────────────
+    ("learning-seam", "candidate teachable-moment seam", "IMPLEMENTED_TESTED",
+     "scripts/research_memory_port.py mark_teachable_moment_candidate", "CANDIDATE evidence via existing channels",
+     "tests/test_memory_port.py", ""),
+    ("learning-seam", "Memory internal-experience port in INWARD", "IMPLEMENTED_TESTED",
+     "scripts/research_memory_port.py; research_inward_context.build_context", "one bounded port of call",
+     "tests/test_memory_port.py", ""),
+    ("learning-seam", "APPLICABILITY governs transfer", "IMPLEMENTED_TESTED",
+     "research_memory_port.applicability_filter + existing APPLICABILITY stage", "irrelevant candidates rejected",
+     "tests/test_memory_port.py::test_applicability_rejects_irrelevant_candidate", ""),
+    ("learning-seam", "zero-result fast path", "IMPLEMENTED_TESTED",
+     "research_memory_port.memory_port_of_call", "[] -> next normal step",
+     "tests/test_memory_port.py; ZERO_RESULT_EXTRA_MODEL_CALLS=0", ""),
+    ("learning-seam", "no competing memory authority", "IMPLEMENTED_TESTED",
+     "port design (provider delegation, no storage)", "Memory stays single future authority",
+     "port creates no store/db; provider contract only", ""),
+
+    # ── DEFERRED (non-V1, with reasons) ────────────────────────────────
+    ("deferred", "automatic Growth Delta pattern learning", "DEFERRED_NON_V1",
+     "src/rune/growth_delta.py (existing, non-V1 scope)", "post-V1 learning loop",
+     "V1 excludes Memory/Substrate/Growth Delta by tranche mandate; the "
+     "T21.5 seam provides the harvest point without implementing the loop", ""),
+    ("deferred", "Memory V1 system itself", "DEFERRED_NON_V1",
+     "(future Memory T0 archaeology)", "dedicated post-V1 phase",
+     "explicitly excluded by T21.5 mandate §0; only the INWARD port is V1", ""),
+    ("deferred", "Substrate", "DEFERRED_NON_V1",
+     "(future)", "post-V1", "explicitly excluded by tranche mandate", ""),
+    ("deferred", "TUI redesign", "DEFERRED_NON_V1",
+     "(existing rune_tui.py adequate)", "explicitly excluded by tranche mandate",
+     "TUI work explicitly out of scope for V1 Scaffolding", ""),
+]
+
+
+def main() -> int:
+    manifest = {
+        "tranche": "T21.5",
+        "rows": [
+            {"family": f, "item": i, "classification": c, "module": m,
+             "authority": a, "evidence": e, "gap": g}
+            for f, i, c, m, a, e, g in ROWS
+        ],
+    }
+    blocking = [r for r in manifest["rows"]
+                if r["classification"] in ("PARTIAL", "ABSENT",
+                                           "IMPLEMENTED_UNPROVEN")]
+    counts: dict[str, int] = {}
+    for r in manifest["rows"]:
+        counts[r["classification"]] = counts.get(r["classification"], 0) + 1
+    manifest["counts"] = counts
+    manifest["blocking_items"] = blocking
+    out = ROOT / "store" / "scaffold" / "v1_parity_manifest.json"
+    out.write_text(json.dumps(manifest, indent=1))
+    print(json.dumps(counts, indent=1))
+    print("blocking mandatory-V1 items:", len(blocking))
+    for b in blocking:
+        print("  BLOCKS:", b["family"], "/", b["item"])
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
@@ -81,6 +81,19 @@ def subjects_since_tag(tag: str, repo: Path | None = None) -> list[str]:
    return out.splitlines() if out else []


+def _version_lt(a: str, b: str) -> bool:
+    """True when semver string a < b. Numeric per segment — never lexicographic.
+
+    T15 Part F: '0.66.9' < '0.66.10' is False lexicographically ('9' > '1'),
+    which made a legitimately-needed patch bump read as the
+    version_ahead_of_target anomaly and silently skip the release
+    (observed live on run 6781, 2026-09-11 19:59).
+    """
+    pa = [int(x) for x in a.split(".")]
+    pb = [int(x) for x in b.split(".")]
+    return pa < pb
+
+
 def classify_subject(subject: str) -> str:
    """Classify one commit subject: ``breaking`` | ``feat`` | ``fix`` | ``other``.

@@ -199,6 +212,218 @@ def decision_in_repo(repo: Path | None = None) -> dict[str, Any]:
    return assemble_release_decision(version, tag, subjects, len(subjects))


+def last_release_boundary(repo: Path | None = None) -> str | None:
+    """Find the last chore(release) commit SHA (the release boundary).
+
+    This is the correct convergence boundary — NOT the git tag (tags may
+    lag behind releases due to D7). Falls back to the latest tag if no
+    release commits exist.
+
+    T15 Part F: matches SUBJECTS only (--grep matches the commit BODY by
+    default; a fix commit whose body mentions 'chore(release)' was
+    misdetected as the boundary, swallowing its own warranted bump).
+    """
+    try:
+        result = subprocess.run(
+            ["git", "log", "--format=%H %s", "-1", "--grep=^chore\\(release\\)", "--extended-regexp"],
+            capture_output=True,
+            text=True,
+            cwd=str(repo) if repo else None,
+            check=True,
+        )
+        for line in result.stdout.splitlines():
+            line = line.strip()
+            if not line:
+                continue
+            sha, _, subject = line.partition(" ")
+            if is_release_commit(subject):
+                return sha
+    except (subprocess.CalledProcessError, IndexError):
+        pass
+    return latest_tag(repo)  # fallback to tag
+
+
+def release_convergence(
+    repo: Path | None = None,
+    *,
+    _version: str | None = None,
+    _boundary_subject: str | None = None,
+    _subjects: list[str] | None = None,
+) -> dict[str, Any]:
+    """Correct release convergence: compare current source version to the
+    warranted TARGET version computed from commits since the last release
+    boundary (the last chore(release) commit, NOT the tag — tags may lag
+    due to D7).
+
+    Invariant:
+      current == target → converged, no commit
+      current < target  → create exactly one bump commit
+      current > target  → anomaly
+    """
+    current_version = _version or read_version(repo)
+    if not current_version:
+        return {
+            "converged": True,
+            "warranted_bump": None,
+            "anomaly": None,
+            "commits_since_tag": 0,
+            "sample": [],
+            "reason": "no readable version",
+        }
+
+    subjects: list[str] = []
+    boundary = None
+    if _subjects is not None:
+        # Test injection: skip git entirely
+        substantive = [s for s in _subjects if not is_release_commit(s)]
+        warranted = highest_warranted_bump(substantive)
+        if not warranted:
+            return {
+                "converged": True,
+                "warranted_bump": None,
+                "anomaly": None,
+                "commits_since_boundary": len(substantive),
+                "sample": substantive[:5],
+                "current_version": current_version,
+                "target_version": current_version,
+                "reason": "no release-worthy commits since last boundary",
+                "boundary_sha": "",
+            }
+        target_version = next_version(current_version, warranted)
+        if current_version == target_version:
+            return {
+                "converged": True,
+                "warranted_bump": warranted,
+                "anomaly": None,
+                "commits_since_boundary": len(substantive),
+                "sample": substantive[:5],
+                "current_version": current_version,
+                "target_version": target_version,
+                "reason": "current version already equals warranted target",
+                "boundary_sha": "",
+            }
+        elif _version_lt(current_version, target_version):
+            return {
+                "converged": False,
+                "warranted_bump": warranted,
+                "anomaly": None,
+                "commits_since_boundary": len(substantive),
+                "sample": substantive[:5],
+                "current_version": current_version,
+                "target_version": target_version,
+                "reason": f"bump required: {current_version} → {target_version}",
+                "boundary_sha": "",
+            }
+        else:
+            return {
+                "converged": False,
+                "warranted_bump": None,
+                "anomaly": "version_ahead_of_target",
+                "commits_since_boundary": len(substantive),
+                "sample": substantive[:5],
+                "current_version": current_version,
+                "target_version": target_version,
+                "reason": f"ANOMALY: current {current_version} > target {target_version}",
+                "boundary_sha": "",
+            }
+    if _boundary_subject:
+        result = subprocess.run(
+            ["git", "log", "--format=%s", f"{boundary}..HEAD"],
+            capture_output=True,
+            text=True,
+            cwd=str(repo) if repo else None,
+            check=True,
+        )
+        subjects = [s for s in result.stdout.splitlines() if s.strip()]
+    else:
+        # Production path: converge over commits since the last release
+        # boundary. T15 Part F root cause: `boundary` was initialized to
+        # None and never assigned — every run classified ALL commits as
+        # unreleased (959+), so the warranted bump never converged and
+        # every queued push wrote its own bump commit (5× 0.66.9 observed
+        # 2026-09-11 20:21). Assign the boundary from the one source of
+        # truth and range the log over it.
+        boundary = last_release_boundary(repo)
+        if boundary:
+            result = subprocess.run(
+                ["git", "log", "--format=%s", f"{boundary}..HEAD"],
+                capture_output=True,
+                text=True,
+                cwd=str(repo) if repo else None,
+                check=True,
+            )
+        else:
+            result = subprocess.run(
+                ["git", "log", "--format=%s"],
+                capture_output=True,
+                text=True,
+                cwd=str(repo) if repo else None,
+                check=True,
+            )
+        subjects = [s for s in result.stdout.splitlines() if s.strip()]
+
+    # Exclude chore(release) commits from bump classification (they ARE releases)
+    substantive = [s for s in subjects if not is_release_commit(s)]
+    warranted = highest_warranted_bump(substantive)
+    # Expose the boundary in the decision JSON: consumers (release step,
+    # debug sessions) otherwise re-derive it from git history.
+    decision_boundary = {"boundary_sha": (boundary or "")[:12]}
+
+    if not warranted:
+        return {
+            "converged": True,
+            "warranted_bump": None,
+            "anomaly": None,
+            "commits_since_boundary": len(substantive),
+            "sample": substantive[:5],
+            "current_version": current_version,
+            "target_version": current_version,
+            "reason": "no release-worthy commits since last boundary",
+            **decision_boundary,
+        }
+
+    target_version = next_version(current_version, warranted)
+
+    if current_version == target_version:
+        return {
+            "converged": True,
+            "warranted_bump": warranted,
+            "anomaly": None,
+            "commits_since_boundary": len(substantive),
+            "sample": substantive[:5],
+            "current_version": current_version,
+            "target_version": target_version,
+            "reason": "current version already equals warranted target",
+            **decision_boundary,
+        }
+    elif _version_lt(current_version, target_version):
+        # Bump needed — but only if not already converged by a prior bump
+        # in this same cascade (the target is the SAME for all queued runs)
+        return {
+            "converged": False,
+            "warranted_bump": warranted,
+            "anomaly": None,
+            "commits_since_boundary": len(substantive),
+            "sample": substantive[:5],
+            "current_version": current_version,
+            "target_version": target_version,
+            "reason": f"bump required: {current_version} → {target_version}",
+            **decision_boundary,
+        }
+    else:
+        return {
+            "converged": False,
+            "warranted_bump": None,
+            "anomaly": "version_ahead_of_target",
+            "commits_since_boundary": len(substantive),
+            "sample": substantive[:5],
+            "current_version": current_version,
+            "target_version": target_version,
+            "reason": f"ANOMALY: current {current_version} > target {target_version}",
+            **decision_boundary,
+        }
+
+
 def main(argv: list[str] | None = None) -> dict[str, Any]:
    parser = argparse.ArgumentParser(description="Release convergence decision (JSON on stdout)")
    parser.add_argument(
@@ -213,7 +438,10 @@ def main(argv: list[str] | None = None) -> dict[str, Any]:
        import sys as _sys

        _sys.exit(0 if is_release_commit(args.is_release_subject) else 1)
-    decision = decision_in_repo(args.repo)
+    decision = release_convergence(args.repo)
+    # Backward compat: old consumers expect commits_since_tag
+    if "commits_since_tag" not in decision:
+        decision["commits_since_tag"] = decision.get("commits_since_boundary", 0)
    print(json.dumps(decision))
    return decision

@@ -0,0 +1,47 @@
+#!/usr/bin/env python3
+"""wr_root_reconcile.py — T21.5 §7: governed split-brain reconciliation.
+
+If canonical and legacy W/R roots both hold live event stores, this
+moves legacy-root events that are ABSENT from the canonical store into
+it (idempotent, content-keyed by filename), then reports what remains.
+Never deletes; the legacy store stays as a moved-marker record.
+"""
+from __future__ import annotations
+
+import json
+import shutil
+import sys
+from pathlib import Path
+
+_SCRIPT_DIR = Path(__file__).resolve().parent
+sys.path.insert(0, str(_SCRIPT_DIR))
+import wr_root_resolver as resolver  # noqa: E402
+
+
+def main() -> int:
+    canonical = resolver.CANONICAL_WR_ROOT
+    legacy = resolver.LEGACY_WR_ROOT
+    (canonical / "events").mkdir(parents=True, exist_ok=True)
+    moved = 0
+    skipped = 0
+    for src in sorted((legacy / "events").glob("*.json")):
+        dst = canonical / "events" / src.name
+        if dst.exists():
+            skipped += 1
+            continue
+        shutil.copy2(src, dst)
+        moved += 1
+    report = {
+        "canonical": str(canonical),
+        "legacy": str(legacy),
+        "events_moved_to_canonical": moved,
+        "already_present_skipped": skipped,
+        "note": "legacy store retained (never deleted); daemon and all "
+                "consumers now resolve via wr_root_resolver",
+    }
+    print(json.dumps(report, indent=2))
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
@@ -0,0 +1,100 @@
+"""wr_root_resolver.py — T21.5 §7: THE ONE wait/resume root authority.
+
+T21.4 live finding: the daemon defaulted WR_ROOT to its own install dir
+while production data lived at /mnt/workdrive/wr-bridge; a restart
+without WR_ROOT exported silently opened a second event store and the
+controller saw no events (the split-brain that forced manual event
+relocation). This module is the SINGLE resolver every consumer shares:
+
+  daemon / event_wait / await_operation tool / notify_event_bridge
+  (CI event writer) / local event writers / CLI / launcher / tests.
+
+Resolution order:
+  1. RUNE_WR_ROOT env (explicit override; tests use it)
+  2. canonical production root (exists?)
+  3. legacy home install root (exists?)
+  4. canonical production root (created on demand — first-run path)
+
+Split-brain policy: if BOTH canonical and legacy roots contain live
+event stores, resolution FAILS LOUDLY unless the env pins one — never
+silently operate two stores. Reconciliation is governed, not silent:
+run scripts/wr_root_reconcile.py.
+"""
+from __future__ import annotations
+
+import json
+import os
+import sys
+from pathlib import Path
+
+CANONICAL_WR_ROOT = Path("/mnt/workdrive/wr-bridge")
+LEGACY_WR_ROOT = Path.home() / ".local" / "share" / "mimo-wait-resume"
+ENV_VAR = "RUNE_WR_ROOT"
+
+SPLIT_BRAIN_ERROR = (
+    "WR ROOT SPLIT-BRAIN: both canonical ({canonical}) and legacy "
+    "({legacy}) contain live event stores. Never operate two stores "
+    "silently — pin {env}=<root> explicitly, or reconcile with "
+    "scripts/wr_root_reconcile.py."
+)
+
+
+def _has_live_store(root: Path) -> bool:
+    """A root counts as 'live' when its event store exists and holds
+    events (an empty dir is just an uninitialized install)."""
+    events = root / "events"
+    if not events.is_dir():
+        return False
+    return any(events.glob("*.json"))
+
+
+def resolve_root(*, require_single: bool = True) -> Path:
+    """Resolve THE wait/resume data root. Raises on split-brain when
+    require_single (default) and no explicit pin exists."""
+    env = os.environ.get(ENV_VAR)
+    if env:
+        return Path(env)
+    canonical_live = _has_live_store(CANONICAL_WR_ROOT)
+    legacy_live = _has_live_store(LEGACY_WR_ROOT)
+    if canonical_live and legacy_live and require_single:
+        raise RuntimeError(SPLIT_BRAIN_ERROR.format(
+            canonical=CANONICAL_WR_ROOT, legacy=LEGACY_WR_ROOT,
+            env=ENV_VAR))
+    if canonical_live or CANONICAL_WR_ROOT.is_dir():
+        return CANONICAL_WR_ROOT
+    if legacy_live or LEGACY_WR_ROOT.is_dir():
+        return LEGACY_WR_ROOT
+    return CANONICAL_WR_ROOT  # first-run: create canonical
+
+
+def daemon_launch_env() -> dict[str, str]:
+    """The env a launcher must pass the daemon: WR_ROOT resolved ONCE
+    from this authority (the daemon itself must not guess)."""
+    return {"WR_ROOT": str(resolve_root()),
+            ENV_VAR: str(resolve_root())}
+
+
+def split_brain_roots() -> list[Path]:
+    """Diagnostic: roots that currently hold live stores."""
+    return [r for r in (CANONICAL_WR_ROOT, LEGACY_WR_ROOT)
+            if _has_live_store(r)]
+
+
+def main() -> int:
+    try:
+        root = resolve_root()
+    except RuntimeError as exc:
+        print(str(exc), file=sys.stderr)
+        return 2
+    print(json.dumps({
+        "resolved_root": str(root),
+        "canonical": str(CANONICAL_WR_ROOT),
+        "legacy": str(LEGACY_WR_ROOT),
+        "split_brain_roots": [str(p) for p in split_brain_roots()],
+        "env_pinned": bool(os.environ.get(ENV_VAR)),
+    }))
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
@@ -352,16 +352,6 @@ def pre_commit_gate_entry(entry: dict) -> dict:
        "critics": critics,
        "reason": "all critics passed" if all_pass else "one or more critics failed"
    }
-    row = db.execute(
-        "SELECT id, text, importance, wash_count, rock_status, topic FROM entries WHERE id = ?",
-        (entry_id,)
-    ).fetchone()
-
-    if not row:
-        return {"verdict": "REJECT", "reason": "entry not found"}
-
-    entry = dict(row)
-    critics = {}

    # Critic 1: Structural validity
    text = entry.get("text", "")
@@ -384,35 +374,6 @@ def pre_commit_gate_entry(entry: dict) -> dict:
    }

    # Critic 3: Semantic consistency (check for contradictions with existing rocks)
-    rocks = db.execute(
-        "SELECT text FROM entries WHERE rock_status = 'rock' AND status = 'active' LIMIT 50"
-    ).fetchall()
-    contradiction_found = False
-    for rock in rocks:
-        rock_text = rock[0].lower()
-        entry_text = text.lower()
-        # Simple contradiction detection: if entry says opposite of a rock
-        if ("always" in rock_text and "never" in entry_text) or \
-           ("never" in rock_text and "always" in entry_text):
-            contradiction_found = True
-            break
-    critics["semantic"] = {
-        "pass": not contradiction_found,
-        "reason": "contradicts existing rock" if contradiction_found else "consistent"
-    }
-
-    # Overall verdict
-    all_pass = all(c["pass"] for c in critics.values())
-    verdict = "PASS" if all_pass else "REJECT"
-
-    return {
-        "verdict": verdict,
-        "entry_id": entry_id,
-        "critics": critics,
-        "reason": "all critics passed" if all_pass else "one or more critics failed"
-    }
-
-
 # ═══════════════════════════════════════════════════════════════════════════════
 # CLI
 # ═══════════════════════════════════════════════════════════════════════════════
@@ -6,7 +6,7 @@ This is the main entry point for the Rune program.
 import json
 from pathlib import Path

-__version__ = "0.66.4"
+__version__ = "0.86.0"

 # One build identity (handoff 2026-09-10 §4): the semantic version above is
 # the repo-level truth (auto-bumped at merge); the DEPLOYED revision (exact
@@ -50,6 +50,7 @@ def _now():
 def connect():
    """Open/create the abstraction database."""
    import sqlite3
+
    os.makedirs(os.path.dirname(DB_PATH), exist_ok=True)
    db = sqlite3.connect(DB_PATH)
    try:
@@ -178,22 +179,24 @@ def llm_call(prompt, model=None, max_tokens=500):
    model = model or LOCAL_MODEL
    try:
        import urllib.request
-        data = json.dumps({
-            "model": model,
-            "prompt": prompt,
-            "stream": False,
-            "options": {"temperature": 0.3, "num_predict": max_tokens, "num_ctx": 8192}
-        }).encode()
+
+        data = json.dumps(
+            {
+                "model": model,
+                "prompt": prompt,
+                "stream": False,
+                "options": {"temperature": 0.3, "num_predict": max_tokens, "num_ctx": 8192},
+            }
+        ).encode()
        req = urllib.request.Request(
-            f"{OLLAMA_URL}/api/generate",
-            data=data,
-            headers={"Content-Type": "application/json"}
+            f"{OLLAMA_URL}/api/generate", data=data, headers={"Content-Type": "application/json"}
        )
        with urllib.request.urlopen(req, timeout=30) as resp:
            result = json.loads(resp.read())
            return result.get("response", "").strip()
    except Exception as e:
        import logging
+
        logging.getLogger("abstraction").warning("LLM call failed: %s", e)
        return ""

@@ -263,8 +266,7 @@ Extract 1-5 key facts. Be precise, not verbose."""
            # Check for duplicates — hash-based (O(1) lookup instead of LIKE scan)
            content_hash = hashlib.sha256(what[:80].lower().encode()).hexdigest()[:16]
            existing = db.execute(
-                "SELECT id FROM facts WHERE state='active' AND content_hash=? LIMIT 1",
-                (content_hash,)
+                "SELECT id FROM facts WHERE state='active' AND content_hash=? LIMIT 1", (content_hash,)
            ).fetchone()
            if existing:
                continue  # NOOP — already exists
@@ -273,12 +275,20 @@ Extract 1-5 key facts. Be precise, not verbose."""
                """INSERT INTO facts (episode_id, what, content_hash, when_text, where_text, who_text, why_text,
                   fact_type, entities_json, confidence, source, state, created_at)
                   VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, 'active', ?)""",
-                (episode_id or "", what, content_hash,
-                 fact.get("when", ""), fact.get("where", ""),
-                 fact.get("who", ""), fact.get("why", ""),
-                 fact.get("type", "world"),
-                 json.dumps(fact.get("entities", [])),
-                 0.7, source, ts)
+                (
+                    episode_id or "",
+                    what,
+                    content_hash,
+                    fact.get("when", ""),
+                    fact.get("where", ""),
+                    fact.get("who", ""),
+                    fact.get("why", ""),
+                    fact.get("type", "world"),
+                    json.dumps(fact.get("entities", [])),
+                    0.7,
+                    source,
+                    ts,
+                ),
            )
            fact_id = cur.lastrowid
            extracted += 1
@@ -291,16 +301,15 @@ Extract 1-5 key facts. Be precise, not verbose."""
                # Upsert entity
                db.execute(
                    "INSERT OR IGNORE INTO entities (name, entity_type, created_at) VALUES (?, 'concept', ?)",
-                    (entity_name, ts)
+                    (entity_name, ts),
                )
                entity_row = db.execute("SELECT id FROM entities WHERE name=?", (entity_name,)).fetchone()
                if entity_row:
                    db.execute(
                        "INSERT OR IGNORE INTO fact_entities (fact_id, entity_id) VALUES (?, ?)",
-                        (fact_id, entity_row["id"])
+                        (fact_id, entity_row["id"]),
                    )
-                    db.execute("UPDATE entities SET fact_count = fact_count + 1 WHERE id=?",
-                               (entity_row["id"],))
+                    db.execute("UPDATE entities SET fact_count = fact_count + 1 WHERE id=?", (entity_row["id"],))

            # Triple extraction (HippoRAG 2 pattern)
            who = fact.get("who", "").strip()
@@ -312,7 +321,7 @@ Extract 1-5 key facts. Be precise, not verbose."""
                    obj = " ".join(words[3:])
                    db.execute(
                        "INSERT INTO triples (subject, predicate, object, source_fact_id, confidence, created_at) VALUES (?, ?, ?, ?, ?, ?)",
-                        (who, predicate, obj, fact_id, 0.6, ts)
+                        (who, predicate, obj, fact_id, 0.6, ts),
                    )

        db.commit()
@@ -335,17 +344,24 @@ def synthesize_observations(entity_name):
        entity_id = entity["id"]

        # Get all facts about this entity
-        facts = db.execute("""
+        facts = db.execute(
+            """
            SELECT f.id, f.what, f.fact_type, f.confidence
            FROM facts f
            JOIN fact_entities fe ON f.id = fe.fact_id
            WHERE fe.entity_id = ? AND f.state = 'active'
            ORDER BY f.created_at DESC
-        """, (entity_id,)).fetchall()
+        """,
+            (entity_id,),
+        ).fetchall()

        if len(facts) < 3:
-            return {"entity": entity_name, "facts": len(facts), "observations": 0,
-                    "reason": "Need at least 3 facts to synthesize"}
+            return {
+                "entity": entity_name,
+                "facts": len(facts),
+                "observations": 0,
+                "reason": "Need at least 3 facts to synthesize",
+            }

        facts_text = "\n".join(f"- [{f['fact_type']}] {f['what']}" for f in facts[:20])

@@ -377,12 +393,11 @@ Respond as JSON array:
            db.execute(
                """INSERT INTO observations (entity_id, observation, source_fact_ids, confidence, state, created_at)
                   VALUES (?, ?, ?, ?, 'active', ?)""",
-                (entity_id, text, json.dumps(source_ids), obs.get("confidence", 0.7), ts)
+                (entity_id, text, json.dumps(source_ids), obs.get("confidence", 0.7), ts),
            )
            created += 1

-        db.execute("UPDATE entities SET observation_count = ? WHERE id=?",
-                   (created, entity_id))
+        db.execute("UPDATE entities SET observation_count = ? WHERE id=?", (created, entity_id))
        db.commit()
        return {"entity": entity_name, "facts": len(facts), "observations": created}
    finally:
@@ -394,8 +409,7 @@ def list_beliefs(limit=20):
    db = connect()
    try:
        rows = db.execute(
-            "SELECT * FROM beliefs WHERE state='active' ORDER BY confidence DESC LIMIT ?",
-            (limit,)
+            "SELECT * FROM beliefs WHERE state='active' ORDER BY confidence DESC LIMIT ?", (limit,)
        ).fetchall()
        return [dict(r) for r in rows]
    finally:
@@ -407,8 +421,7 @@ def list_rules(limit=20):
    db = connect()
    try:
        rows = db.execute(
-            "SELECT * FROM rules WHERE state='active' ORDER BY confidence DESC LIMIT ?",
-            (limit,)
+            "SELECT * FROM rules WHERE state='active' ORDER BY confidence DESC LIMIT ?", (limit,)
        ).fetchall()
        return [dict(r) for r in rows]
    finally:
@@ -420,8 +433,7 @@ def list_skills(limit=20):
    db = connect()
    try:
        rows = db.execute(
-            "SELECT * FROM skills WHERE state='active' ORDER BY success_count DESC LIMIT ?",
-            (limit,)
+            "SELECT * FROM skills WHERE state='active' ORDER BY success_count DESC LIMIT ?", (limit,)
        ).fetchall()
        return [dict(r) for r in rows]
    finally:
@@ -486,8 +498,9 @@ def handle_hook(event, data=None):

 def main():
    ap = argparse.ArgumentParser(description="Abstraction pipeline")
-    ap.add_argument("cmd", choices=["extract", "observe", "beliefs", "rules", "skills",
-                                     "evolve", "lifecycle", "stats", "hook"])
+    ap.add_argument(
+        "cmd", choices=["extract", "observe", "beliefs", "rules", "skills", "evolve", "lifecycle", "stats", "hook"]
+    )
    ap.add_argument("arg", nargs="?", default="")
    ap.add_argument("--episode-id", default=None)
    ap.add_argument("--source", default="session")
@@ -501,16 +514,18 @@ def main():
            print("usage: abstraction.py extract '<episode text>'")
            return
        result = extract_facts(args.arg, episode_id=args.episode_id, source=args.source)
-        print(json.dumps(result, indent=2) if args.json else
-              f"Extracted {result.get('extracted', 0)} facts")
+        print(json.dumps(result, indent=2) if args.json else f"Extracted {result.get('extracted', 0)} facts")

    elif args.cmd == "observe":
        if not args.arg:
            print("usage: abstraction.py observe '<entity name>'")
            return
        result = synthesize_observations(args.arg)
-        print(json.dumps(result, indent=2) if args.json else
-              f"Entity: {result.get('entity')} — {result.get('observations', 0)} observations from {result.get('facts', 0)} facts")
+        print(
+            json.dumps(result, indent=2)
+            if args.json
+            else f"Entity: {result.get('entity')} — {result.get('observations', 0)} observations from {result.get('facts', 0)} facts"
+        )

    elif args.cmd == "beliefs":
        items = list_beliefs()
@@ -568,5 +583,5 @@ PLUGIN = {
    "dependencies": [],
    "provides": ["abstraction"],
    "priority": 50,
-    "auto_activate": True
+    "auto_activate": True,
 }
@@ -562,7 +562,25 @@ def derive_sufficiency(
    # appears. A leading inspection verb makes the task read-only.
    first_words = lower.split()[:3]
    if any(
-        w in ("inspect", "read", "review", "analyze", "analyse", "explain", "look", "list", "show", "describe", "check", "what", "why", "how", "where", "which")
+        w
+        in (
+            "inspect",
+            "read",
+            "review",
+            "analyze",
+            "analyse",
+            "explain",
+            "look",
+            "list",
+            "show",
+            "describe",
+            "check",
+            "what",
+            "why",
+            "how",
+            "where",
+            "which",
+        )
        for w in first_words
    ):
        is_defect = False
@@ -30,6 +30,7 @@ if _SD not in sys.path:
@dataclass
 class AttentionCapsule:
    """Compact working set for cognitive work."""
+
    active_objective: str = ""
    current_phase: str = ""
    relevant_memories: List[str] = field(default_factory=list)
@@ -199,14 +200,14 @@ def create_capsule(task: str, context: str = "", max_tokens: int = 600) -> Atten

    # Extract task requirements (not constraints — no refusal language)
    requirement_patterns = [
-        r'need to',
-        r'want to',
-        r'please',
-        r'fix',
-        r'build',
-        r'create',
-        r'update',
-        r'change',
+        r"need to",
+        r"want to",
+        r"please",
+        r"fix",
+        r"build",
+        r"create",
+        r"update",
+        r"change",
    ]

    for pattern in requirement_patterns:
@@ -233,7 +234,7 @@ def create_capsule(task: str, context: str = "", max_tokens: int = 600) -> Atten

        # If still over, trim objective
        while capsule.token_count > max_tokens and len(capsule.active_objective) > 50:
-            capsule.active_objective = capsule.active_objective[:len(capsule.active_objective)//2]
+            capsule.active_objective = capsule.active_objective[: len(capsule.active_objective) // 2]
            prompt_text = capsule.to_prompt()
            capsule.token_count = estimate_tokens(prompt_text)

@@ -291,5 +292,5 @@ PLUGIN = {
    "dependencies": [],
    "provides": ["attention_capsule"],
    "priority": 5,
-    "auto_activate": True
+    "auto_activate": True,
 }
@@ -87,14 +87,14 @@ def verify_chain() -> dict:
        try:
            entry = json.loads(line)
        except json.JSONDecodeError:
-            return {"valid": False, "entries": i, "reason": f"invalid JSON at line {i+1}"}
+            return {"valid": False, "entries": i, "reason": f"invalid JSON at line {i + 1}"}

        # Verify prev_hash chain
        if entry.get("prev_hash") != prev_hash:
            return {
                "valid": False,
                "entries": i,
-                "reason": f"chain broken at line {i+1}: expected prev_hash={prev_hash}, got {entry.get('prev_hash')}"
+                "reason": f"chain broken at line {i + 1}: expected prev_hash={prev_hash}, got {entry.get('prev_hash')}",
            }

        # Verify HMAC
@@ -104,7 +104,7 @@ def verify_chain() -> dict:
            return {
                "valid": False,
                "entries": i,
-                "reason": f"HMAC mismatch at line {i+1}: stored={stored_hash[:16]}..., computed={computed[:16]}..."
+                "reason": f"HMAC mismatch at line {i + 1}: stored={stored_hash[:16]}..., computed={computed[:16]}...",
            }

        prev_hash = stored_hash
@@ -131,37 +131,51 @@ def get_entries(limit: int = 50) -> List[dict]:

 def log_promotion(entry_id: int, old_status: str, new_status: str, wash_count: int, reason: str) -> dict:
    """Log a promotion decision."""
-    return append_entry("promotion", entry_id, {
-        "old_status": old_status,
-        "new_status": new_status,
-        "wash_count": wash_count,
-        "reason": reason,
-    })
+    return append_entry(
+        "promotion",
+        entry_id,
+        {
+            "old_status": old_status,
+            "new_status": new_status,
+            "wash_count": wash_count,
+            "reason": reason,
+        },
+    )


 def log_gate_result(entry_id: int, verdict: str, critics: dict) -> dict:
    """Log a pre-commit gate result."""
-    return append_entry("gate", entry_id, {
-        "verdict": verdict,
-        "critics": critics,
-    })
+    return append_entry(
+        "gate",
+        entry_id,
+        {
+            "verdict": verdict,
+            "critics": critics,
+        },
+    )


 def log_drift_alert(alert_type: str, message: str, severity: str) -> dict:
    """Log a drift alert."""
-    return append_entry("alert", 0, {
-        "alert_type": alert_type,
-        "message": message,
-        "severity": severity,
-    })
+    return append_entry(
+        "alert",
+        0,
+        {
+            "alert_type": alert_type,
+            "message": message,
+            "severity": severity,
+        },
+    )


 # ═══════════════════════════════════════════════════════════════════════════════
 # CLI
 # ═══════════════════════════════════════════════════════════════════════════════

+
 def main():
    import argparse
+
    parser = argparse.ArgumentParser(description="Promotion audit ledger")
    sub = parser.add_subparsers(dest="cmd")

@@ -198,10 +212,12 @@ def main():
            print(f"Recent ledger entries ({len(entries)}):")
            for e in entries:
                print(f"  [{e.get('type')}] entry_id={e.get('entry_id')} @ {e.get('timestamp', '')[:19]}")
-                details = e.get('details', {})
-                if e.get('type') == 'promotion':
-                    print(f"    {details.get('old_status')} → {details.get('new_status')} (wash={details.get('wash_count')})")
-                elif e.get('type') == 'gate':
+                details = e.get("details", {})
+                if e.get("type") == "promotion":
+                    print(
+                        f"    {details.get('old_status')} → {details.get('new_status')} (wash={details.get('wash_count')})"
+                    )
+                elif e.get("type") == "gate":
                    print(f"    verdict={details.get('verdict')}")
        else:
            print("No ledger entries.")
@@ -28,8 +28,7 @@ def run_warm_router(message):
    """Run warm router and return fired neurons."""
    try:
        result = subprocess.run(
-            [PYTHON, WARM_ROUTER, "route", message, "--json"],
-            capture_output=True, text=True, timeout=10
+            [PYTHON, WARM_ROUTER, "route", message, "--json"], capture_output=True, text=True, timeout=10
        )
        if result.returncode == 0 and result.stdout.strip():
            return json.loads(result.stdout)
@@ -42,8 +41,7 @@ def run_memory_recall(query, budget=5):
    """Run memory recall and return relevant entries."""
    try:
        result = subprocess.run(
-            [PYTHON, MEMORY_STORE, "recall", query, "--json"],
-            capture_output=True, text=True, timeout=15
+            [PYTHON, MEMORY_STORE, "recall", query, "--json"], capture_output=True, text=True, timeout=15
        )
        if result.returncode == 0 and result.stdout.strip():
            data = json.loads(result.stdout)
@@ -57,10 +55,7 @@ def run_memory_recall(query, budget=5):
 def run_always_digest():
    """Run always-on digest and return results."""
    try:
-        result = subprocess.run(
-            [PYTHON, WARM_ROUTER, "always", "--json"],
-            capture_output=True, text=True, timeout=10
-        )
+        result = subprocess.run([PYTHON, WARM_ROUTER, "always", "--json"], capture_output=True, text=True, timeout=10)
        if result.returncode == 0 and result.stdout.strip():
            return json.loads(result.stdout)
    except Exception:
@@ -108,14 +103,70 @@ def build_auto_context(message):
    # Memory recall for substantive messages
    if message and len(message.strip()) > 10:
        # Extract key terms for recall (words > 4 chars, not stopwords)
-        stopwords = {"what", "when", "where", "which", "that", "this", "with", "from",
-                     "have", "been", "were", "they", "their", "about", "would", "could",
-                     "should", "will", "into", "your", "just", "like", "know", "think",
-                     "also", "back", "then", "than", "more", "some", "only", "very",
-                     "does", "done", "make", "made", "want", "need", "give", "tell",
-                     "there", "these", "those", "because", "between", "after", "before",
-                     "under", "over", "again", "once", "here", "each", "every", "both",
-                     "few", "most", "other", "such", "same", "being", "having"}
+        stopwords = {
+            "what",
+            "when",
+            "where",
+            "which",
+            "that",
+            "this",
+            "with",
+            "from",
+            "have",
+            "been",
+            "were",
+            "they",
+            "their",
+            "about",
+            "would",
+            "could",
+            "should",
+            "will",
+            "into",
+            "your",
+            "just",
+            "like",
+            "know",
+            "think",
+            "also",
+            "back",
+            "then",
+            "than",
+            "more",
+            "some",
+            "only",
+            "very",
+            "does",
+            "done",
+            "make",
+            "made",
+            "want",
+            "need",
+            "give",
+            "tell",
+            "there",
+            "these",
+            "those",
+            "because",
+            "between",
+            "after",
+            "before",
+            "under",
+            "over",
+            "again",
+            "once",
+            "here",
+            "each",
+            "every",
+            "both",
+            "few",
+            "most",
+            "other",
+            "such",
+            "same",
+            "being",
+            "having",
+        }
        words = [w for w in message.lower().split() if len(w) > 4 and w.isalpha() and w not in stopwords]
        if words:
            query = " ".join(words[:5])
@@ -167,6 +218,7 @@ def handle_hook(event, data=None):

 def main():
    import argparse
+
    ap = argparse.ArgumentParser(description="Auto-context: automatic memory activation")
    ap.add_argument("cmd", choices=["build", "show", "hook"])
    ap.add_argument("arg", nargs="?", default="")
@@ -214,5 +266,5 @@ PLUGIN = {
    "dependencies": ["memory_store", "warm_router"],
    "provides": ["auto_context"],
    "priority": 5,
-    "auto_activate": True
+    "auto_activate": True,
 }
@@ -85,63 +85,72 @@ def extract_tasks(message, channel="chat"):
    """
    # Strip code blocks before analysis
    import re as _re
-    message = _re.sub(r'```[\s\S]*?```', '', message)
-    message = _re.sub(r'`[^`]+`', '', message)
+
+    message = _re.sub(r"```[\s\S]*?```", "", message)
+    message = _re.sub(r"`[^`]+`", "", message)

    tasks = []
    lines = message.split("\n")

    # Pattern 1: Numbered lists (1. do this, 2. do that)
-    numbered = re.findall(r'(?:^|\n)\s*(\d+)[\.\)]\s*(.+?)(?:\n|$)', message)
+    numbered = re.findall(r"(?:^|\n)\s*(\d+)[\.\)]\s*(.+?)(?:\n|$)", message)
    if len(numbered) >= 2:
        for num, text in numbered:
            text = text.strip()
            if len(text) > 10:
-                tasks.append({
-                    "description": text,
-                    "priority": "normal",
-                    "pattern": "numbered_list",
-                })
+                tasks.append(
+                    {
+                        "description": text,
+                        "priority": "normal",
+                        "pattern": "numbered_list",
+                    }
+                )

    # Pattern 2: Bullet points (- do this, * do that)
-    bullets = re.findall(r'(?:^|\n)\s*[-*•]\s*(.+?)(?:\n|$)', message)
+    bullets = re.findall(r"(?:^|\n)\s*[-*•]\s*(.+?)(?:\n|$)", message)
    if len(bullets) >= 2:
        for text in bullets:
            text = text.strip()
            if len(text) > 10:
-                tasks.append({
-                    "description": text,
-                    "priority": "normal",
-                    "pattern": "bullet_list",
-                })
+                tasks.append(
+                    {
+                        "description": text,
+                        "priority": "normal",
+                        "pattern": "bullet_list",
+                    }
+                )

    # Pattern 3: Action verbs at start of sentences
    action_patterns = [
-        r'(?:^|\n)\s*(?:I(?:\'ll| will| need to)|let me|we should|we need to)\s+(.+?)(?:\.|$|\n)',
-        r'(?:^|\n)\s*(?:fix|build|create|add|update|check|verify|test|research|investigate|implement|design)\s+(.+?)(?:\.|$|\n)',
+        r"(?:^|\n)\s*(?:I(?:\'ll| will| need to)|let me|we should|we need to)\s+(.+?)(?:\.|$|\n)",
+        r"(?:^|\n)\s*(?:fix|build|create|add|update|check|verify|test|research|investigate|implement|design)\s+(.+?)(?:\.|$|\n)",
    ]
    for pattern in action_patterns:
        matches = re.findall(pattern, message, re.IGNORECASE | re.MULTILINE)
        for text in matches:
            text = text.strip()
            if len(text) > 10 and len(text) < 200:
-                tasks.append({
-                    "description": text,
-                    "priority": "normal",
-                    "pattern": "action_verb",
-                })
+                tasks.append(
+                    {
+                        "description": text,
+                        "priority": "normal",
+                        "pattern": "action_verb",
+                    }
+                )

    # Pattern 4: "First... then... finally..." sequences
-    sequence = re.findall(r'(?:first|then|next|finally|after that)[,.]?\s+(.+?)(?:\.|$|\n)', message, re.IGNORECASE)
+    sequence = re.findall(r"(?:first|then|next|finally|after that)[,.]?\s+(.+?)(?:\.|$|\n)", message, re.IGNORECASE)
    if len(sequence) >= 2:
        for text in sequence:
            text = text.strip()
            if len(text) > 10:
-                tasks.append({
-                    "description": text,
-                    "priority": "normal",
-                    "pattern": "sequence",
-                })
+                tasks.append(
+                    {
+                        "description": text,
+                        "priority": "normal",
+                        "pattern": "sequence",
+                    }
+                )

    # Deduplicate by description similarity
    seen = set()
@@ -168,12 +177,14 @@ def store_tasks(tasks, channel="chat", source_message=""):
            next_id = (row["max_id"] or 0) + 1
            task_id = f"AT{next_id}"

-            db.execute("""
+            db.execute(
+                """
                INSERT INTO tasks (task_id, description, source_message, source_channel,
                                  status, priority, created_at, updated_at)
                VALUES (?, ?, ?, ?, 'active', ?, ?, ?)
-            """, (task_id, task["description"], source_message[:500], channel,
-                  task.get("priority", "normal"), ts, ts))
+            """,
+                (task_id, task["description"], source_message[:500], channel, task.get("priority", "normal"), ts, ts),
+            )

            # Log to completion file
            log_completion(task_id, "created", task["description"], channel)
@@ -196,15 +207,21 @@ def complete_task(task_id, notes=""):
        row = db.execute("SELECT description FROM tasks WHERE task_id=? AND status='active'", (task_id,)).fetchone()
        description = row["description"] if row else ""

-        db.execute("""
+        db.execute(
+            """
            UPDATE tasks SET status='completed', completion_notes=?, completed_at=?, updated_at=?
            WHERE task_id=? AND status='active'
-        """, (notes, ts, ts, task_id))
+        """,
+            (notes, ts, ts, task_id),
+        )

-        db.execute("""
+        db.execute(
+            """
            INSERT INTO completion_log (task_id, action, details, timestamp)
            VALUES (?, 'completed', ?, ?)
-        """, (task_id, notes, ts))
+        """,
+            (task_id, notes, ts),
+        )

        db.commit()
    finally:
@@ -244,6 +261,7 @@ def run_auto_checks(description):
 def run_compile_check():
    """Check that all Python scripts compile."""
    import subprocess
+
    try:
        check_script = (
            "import py_compile, glob, sys; "
@@ -253,9 +271,7 @@ def run_compile_check():
            "print(f'FAIL:{len(failed)}' if failed else 'PASS')"
        )
        result = subprocess.run(
-            [sys.executable, "-c", check_script],
-            capture_output=True, text=True, timeout=30,
-            cwd=SCRIPTS_DIR
+            [sys.executable, "-c", check_script], capture_output=True, text=True, timeout=30, cwd=SCRIPTS_DIR
        )
        passed = "PASS" in result.stdout
        return {"check": "compile", "passed": passed, "details": result.stdout.strip()[:200]}
@@ -266,14 +282,18 @@ def run_compile_check():
 def run_benchmark_check():
    """Run RuneBench."""
    import subprocess
+
    try:
        result = subprocess.run(
            [sys.executable, os.path.join(SCRIPTS_DIR, "rune_bench.py"), "run"],
-            capture_output=True, text=True, timeout=120
+            capture_output=True,
+            text=True,
+            timeout=120,
        )
        # Check for pass patterns: "N/N" where both numbers match, or explicit ALL PASS
        import re
-        pass_match = re.search(r'(\d+)/\1', result.stdout)
+
+        pass_match = re.search(r"(\d+)/\1", result.stdout)
        passed = bool(pass_match) or "ALL" in result.stdout.upper() or "PASS" in result.stdout.upper()
        return {"check": "benchmark", "passed": passed, "details": result.stdout.strip()[-200:]}
    except Exception as e:
@@ -284,12 +304,17 @@ def run_sanity_check():
    """Basic sanity check — can we import key modules?"""
    try:
        import subprocess
+
        result = subprocess.run(
-            [sys.executable, "-c",
-             "import sys; sys.path.insert(0, '.'); "
-             "import memory_store, warm_router, microkernel; print('PASS')"],
-            capture_output=True, text=True, timeout=10,
-            cwd=SCRIPTS_DIR
+            [
+                sys.executable,
+                "-c",
+                "import sys; sys.path.insert(0, '.'); import memory_store, warm_router, microkernel; print('PASS')",
+            ],
+            capture_output=True,
+            text=True,
+            timeout=10,
+            cwd=SCRIPTS_DIR,
        )
        passed = "PASS" in result.stdout
        return {"check": "sanity", "passed": passed, "details": result.stdout.strip()[:200]}
@@ -302,8 +327,7 @@ def list_tasks(status="active", limit=20):
    db = connect()
    try:
        rows = db.execute(
-            "SELECT * FROM tasks WHERE status=? ORDER BY created_at DESC LIMIT ?",
-            (status, limit)
+            "SELECT * FROM tasks WHERE status=? ORDER BY created_at DESC LIMIT ?", (status, limit)
        ).fetchall()
        return [dict(r) for r in rows]
    finally:
@@ -359,6 +383,7 @@ def handle_hook(event, data=None):

 def main():
    import argparse
+
    ap = argparse.ArgumentParser(description="Automatic task extraction")
    ap.add_argument("cmd", choices=["extract", "complete", "list", "completion", "hook"])
    ap.add_argument("arg", nargs="?", default="")
@@ -431,5 +456,5 @@ PLUGIN = {
    "dependencies": [],
    "provides": ["auto_tasks"],
    "priority": 4,
-    "auto_activate": True
+    "auto_activate": True,
 }
@@ -101,11 +101,14 @@ def add_correction(original, corrected, scope_tags=None, domain="", pattern=""):
        ts = _now()
        tags = json.dumps(scope_tags or {})

-        cur = db.execute("""
+        cur = db.execute(
+            """
            INSERT INTO corrections (original_behavior, corrected_behavior, scope_tags,
               scope_pattern, scope_domain, confidence, state, created_at)
            VALUES (?, ?, ?, ?, ?, 0.5, 'active', ?)
-        """, (original, corrected, tags, pattern, domain, ts))
+        """,
+            (original, corrected, tags, pattern, domain, ts),
+        )
        correction_id = cur.lastrowid
        db.commit()
        return {"added": correction_id, "confidence": 0.5, "scope": scope_tags or {}}
@@ -120,11 +123,14 @@ def add_habit(trigger, action, name="", context_tags=None, min_activations=3):
        ts = _now()
        tags = json.dumps(context_tags or [])

-        cur = db.execute("""
+        cur = db.execute(
+            """
            INSERT INTO habits (name, trigger_pattern, context_tags, action_template,
               min_activations, confidence, state, created_at)
            VALUES (?, ?, ?, ?, ?, 0.5, 'candidate', ?)
-        """, (name, trigger, tags, action, min_activations, ts))
+        """,
+            (name, trigger, tags, action, min_activations, ts),
+        )
        habit_id = cur.lastrowid
        db.commit()
        return {"added": habit_id, "state": "candidate", "min_activations": min_activations}
@@ -214,16 +220,22 @@ def reinforce(item_id, outcome, item_type="habit"):
            if habit["state"] == "candidate" and new_count >= habit["min_activations"]:
                new_state = "active"

-            db.execute("""
+            db.execute(
+                """
                UPDATE habits SET activation_count=?, success_count=?, failure_count=?,
                confidence=?, state=?, last_activated=? WHERE id=?
-            """, (new_count, new_success, new_failure, new_confidence, new_state, ts, item_id))
+            """,
+                (new_count, new_success, new_failure, new_confidence, new_state, ts, item_id),
+            )

            # Log activation
-            db.execute("""
+            db.execute(
+                """
                INSERT INTO habit_activations (habit_id, context, outcome, activated_at)
                VALUES (?, '', ?, ?)
-            """, (item_id, outcome, ts))
+            """,
+                (item_id, outcome, ts),
+            )

            db.commit()
            return {"reinforced": item_id, "confidence": new_confidence, "state": new_state}
@@ -244,10 +256,13 @@ def reinforce(item_id, outcome, item_type="habit"):
                new_count = correction["evidence_count"]

            new_state = "active" if new_confidence > 0.3 else "archived"
-            db.execute("""
+            db.execute(
+                """
                UPDATE corrections SET confidence=?, evidence_count=?, state=?, last_reinforced=?
                WHERE id=?
-            """, (new_confidence, new_count, new_state, ts, item_id))
+            """,
+                (new_confidence, new_count, new_state, ts, item_id),
+            )
            db.commit()
            return {"reinforced": item_id, "confidence": new_confidence, "state": new_state}
    finally:
@@ -261,17 +276,23 @@ def list_items(item_type="all", limit=20):
        results = {}

        if item_type in ("all", "correction"):
-            corrections = db.execute("""
+            corrections = db.execute(
+                """
                SELECT * FROM corrections WHERE state='active'
                ORDER BY confidence DESC LIMIT ?
-            """, (limit,)).fetchall()
+            """,
+                (limit,),
+            ).fetchall()
            results["corrections"] = [dict(c) for c in corrections]

        if item_type in ("all", "habit"):
-            habits = db.execute("""
+            habits = db.execute(
+                """
                SELECT * FROM habits WHERE state IN ('active', 'candidate')
                ORDER BY confidence DESC LIMIT ?
-            """, (limit,)).fetchall()
+            """,
+                (limit,),
+            ).fetchall()
            results["habits"] = [dict(h) for h in habits]

        return results
@@ -285,7 +306,9 @@ def behavior_stats():
    try:
        stats = {}
        stats["corrections_active"] = db.execute("SELECT COUNT(*) FROM corrections WHERE state='active'").fetchone()[0]
-        stats["corrections_archived"] = db.execute("SELECT COUNT(*) FROM corrections WHERE state='archived'").fetchone()[0]
+        stats["corrections_archived"] = db.execute(
+            "SELECT COUNT(*) FROM corrections WHERE state='archived'"
+        ).fetchone()[0]
        stats["habits_active"] = db.execute("SELECT COUNT(*) FROM habits WHERE state='active'").fetchone()[0]
        stats["habits_candidate"] = db.execute("SELECT COUNT(*) FROM habits WHERE state='candidate'").fetchone()[0]
        stats["total_activations"] = db.execute("SELECT COUNT(*) FROM habit_activations").fetchone()[0]
@@ -296,6 +319,7 @@ def behavior_stats():

 def main():
    import argparse
+
    ap = argparse.ArgumentParser(description="Behavioral patterns")
    ap.add_argument("cmd", choices=["correct", "habit", "match", "reinforce", "list", "stats"])
    ap.add_argument("arg", nargs="?", default="")
@@ -317,20 +341,16 @@ def main():
            print("usage: behavior.py correct '<original>' '<corrected>'")
            return
        scope = json.loads(args.scope) if args.scope else None
-        result = add_correction(args.arg, args.arg2, scope_tags=scope,
-                               domain=args.domain, pattern=args.pattern)
-        print(json.dumps(result, indent=2) if args.json else
-              f"Correction added (confidence: {result['confidence']})")
+        result = add_correction(args.arg, args.arg2, scope_tags=scope, domain=args.domain, pattern=args.pattern)
+        print(json.dumps(result, indent=2) if args.json else f"Correction added (confidence: {result['confidence']})")

    elif args.cmd == "habit":
        if not all([args.arg, args.arg2]):
            print("usage: behavior.py habit '<trigger_pattern>' '<action>'")
            return
        tags = json.loads(args.context_tags) if args.context_tags else None
-        result = add_habit(args.arg, args.arg2, name=args.name,
-                          context_tags=tags, min_activations=args.min_activations)
-        print(json.dumps(result, indent=2) if args.json else
-              f"Habit candidate added (state: {result['state']})")
+        result = add_habit(args.arg, args.arg2, name=args.name, context_tags=tags, min_activations=args.min_activations)
+        print(json.dumps(result, indent=2) if args.json else f"Habit candidate added (state: {result['state']})")

    elif args.cmd == "match":
        if not args.arg:
@@ -350,8 +370,7 @@ def main():
            print("usage: behavior.py reinforce <id>")
            return
        result = reinforce(int(args.arg), args.outcome, item_type=args.type)
-        print(json.dumps(result, indent=2) if args.json else
-              f"Reinforced: confidence={result.get('confidence', '?')}")
+        print(json.dumps(result, indent=2) if args.json else f"Reinforced: confidence={result.get('confidence', '?')}")

    elif args.cmd == "list":
        results = list_items(item_type=args.type, limit=args.limit)
@@ -361,15 +380,18 @@ def main():
            for c in results.get("corrections", []):
                print(f"  [correction {c['id']}] {c['corrected_behavior'][:60]} (conf: {c['confidence']:.2f})")
            for h in results.get("habits", []):
-                state = h['state']
+                state = h["state"]
                print(f"  [habit {h['id']}] {h['action_template'][:60]} ({state}, conf: {h['confidence']:.2f})")

    elif args.cmd == "stats":
        s = behavior_stats()
-        print(json.dumps(s, indent=2) if args.json else
-              f"Corrections: {s['corrections_active']} active, {s['corrections_archived']} archived | "
-              f"Habits: {s['habits_active']} active, {s['habits_candidate']} candidate | "
-              f"Activations: {s['total_activations']}")
+        print(
+            json.dumps(s, indent=2)
+            if args.json
+            else f"Corrections: {s['corrections_active']} active, {s['corrections_archived']} archived | "
+            f"Habits: {s['habits_active']} active, {s['habits_candidate']} candidate | "
+            f"Activations: {s['total_activations']}"
+        )


 if __name__ == "__main__":
@@ -385,5 +407,5 @@ PLUGIN = {
    "dependencies": [],
    "provides": ["behavior"],
    "priority": 50,
-    "auto_activate": True
+    "auto_activate": True,
 }
@@ -0,0 +1,251 @@
+"""behaviour_ledger.py — T19 §19F/19G/19I: the canonical Behaviour
+Validation Ledger.
+
+PRIMARY UNIT: the VERSIONED BEHAVIOUR / TEST SPECIFICATION — not "a
+test run in some session". Automated verification and real-world/user
+confirmation are DISTINCT truths (rule 19F): tests green ≠ user-
+confirmed.
+
+GENERATION-AWARE HISTORY (rule 19G, mandatory): when a dependency
+changes, affected behaviours become RETEST_REQUIRED while their prior
+confirmation history stays queryable — confidence does not silently
+carry forward, and it does not silently disappear either.
+
+SELECTIVE INVALIDATION (rule 19I): only behaviours whose structural
+dependencies changed are invalidated; every invalidation carries a
+reason; unrelated hardened confirmations are preserved.
+
+Storage: <project_store>/behaviour_ledger.json (versioned document).
+"""
+from __future__ import annotations
+
+import time
+from pathlib import Path
+
+LEDGER_SCHEMA_VERSION = 1
+
+# hardening lifecycle (rule 19G semantics)
+UNVALIDATED = "UNVALIDATED"                    # not yet validated
+IMPLEMENTED = "IMPLEMENTED"                    # implementation exists
+AUTO_VERIFIED = "AUTO_VERIFIED"                # automated verification passed
+REAL_WORLD_PASSED = "REAL_WORLD_PASSED"        # manual/real test passed
+TEST_FAILED = "TEST_FAILED"                    # test failed
+PARTIAL = "PARTIAL"                            # inconclusive
+USER_CONFIRMED = "USER_CONFIRMED"              # hardened
+RETEST_REQUIRED = "RETEST_REQUIRED"            # confidence invalidated
+ACCEPTED_LIMITATION = "ACCEPTED_LIMITATION"    # explicitly approved gap
+NOT_APPLICABLE = "NOT_APPLICABLE"
+
+STATES = (UNVALIDATED, IMPLEMENTED, AUTO_VERIFIED, REAL_WORLD_PASSED,
+          TEST_FAILED, PARTIAL, USER_CONFIRMED, RETEST_REQUIRED,
+          ACCEPTED_LIMITATION, NOT_APPLICABLE)
+
+# states that represent STANDING CONFIDENCE (selective invalidation
+# may reopen exactly these; everything else is already open or n/a)
+CONFIRMED_STATES = (AUTO_VERIFIED, REAL_WORLD_PASSED, USER_CONFIRMED)
+
+
+def _now() -> str:
+    return time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime())
+
+
+class BehaviourLedger:
+    def __init__(self, store_dir: Path | str):
+        self.path = Path(store_dir) / "behaviour_ledger.json"
+        self.doc = self._load()
+
+    def _load(self) -> dict:
+        if self.path.is_file():
+            import json
+            return json.loads(self.path.read_text())
+        return {"schema_version": LEDGER_SCHEMA_VERSION,
+                "behaviours": {}, "updated_at": _now()}
+
+    def save(self) -> None:
+        import json
+        self.doc["updated_at"] = _now()
+        self.path.parent.mkdir(parents=True, exist_ok=True)
+        self.path.write_text(json.dumps(self.doc, indent=1, sort_keys=True))
+
+    # -- specification ------------------------------------------------------
+
+    def define(self, *, behaviour_id: str, owning_project: str,
+               intended_behaviour: str, origin_intent: str = "",
+               pass_criteria: str = "", fail_criteria: str = "",
+               procedure: str = "",
+               owner_component: str = "",
+               dependencies: list[str] | None = None,
+               automated_ref: str = "",
+               environment: str = "") -> dict:
+        bid = (behaviour_id if behaviour_id.startswith("behaviour:")
+               else f"behaviour:{behaviour_id}")
+        if bid in self.doc["behaviours"]:
+            raise ValueError(f"behaviour exists: {bid}")
+        rec = {
+            "behaviour_id": bid,
+            "owning_project": owning_project,
+            "intended_behaviour": intended_behaviour,
+            "origin_intent": origin_intent,
+            "pass_criteria": pass_criteria,
+            "fail_criteria": fail_criteria,
+            "procedure": procedure,
+            "owner_component": owner_component,   # structural entity id
+            "dependencies": list(dependencies or []),  # structural ids
+            "automated_ref": automated_ref,
+            "environment": environment,
+            "state": UNVALIDATED,
+            "evidence": [],           # [{at, generation, kind, detail}]
+            "history": [],            # every state transition, never lost
+            "invalidations": [],      # [{at, generation, because}]
+            "limitations": [],
+            "created_at": _now(),
+            "updated_at": _now(),
+        }
+        self.doc["behaviours"][bid] = rec
+        return rec
+
+    def get(self, behaviour_id: str) -> dict | None:
+        return self.doc["behaviours"].get(
+            behaviour_id if behaviour_id.startswith("behaviour:")
+            else f"behaviour:{behaviour_id}")
+
+    # -- state transitions (all recorded in history — rule 19G) -------------
+
+    def _transition(self, rec: dict, new_state: str, *, kind: str,
+                    detail: str = "", generation: str = "") -> None:
+        if new_state not in STATES:
+            raise ValueError(f"unknown validation state {new_state!r}")
+        rec["history"].append({
+            "at": _now(), "from": rec["state"], "to": new_state,
+            "kind": kind, "detail": detail[:200],
+            "generation": generation})
+        rec["state"] = new_state
+        rec["updated_at"] = _now()
+
+    def record_auto_verification(self, behaviour_id: str, *,
+                                 passed: bool, detail: str = "",
+                                 generation: str = "") -> dict:
+        """Automated verification. GREEN NEVER IMPLIES USER_CONFIRMED
+        (rule 19W-22): the ceiling here is AUTO_VERIFIED."""
+        rec = self.get(behaviour_id)
+        if rec is None:
+            raise ValueError(f"unknown behaviour {behaviour_id!r}")
+        rec["evidence"].append({"at": _now(), "kind": "automated",
+                                "passed": passed,
+                                "detail": detail[:200],
+                                "generation": generation})
+        self._transition(rec, AUTO_VERIFIED if passed else TEST_FAILED,
+                         kind="automated_verification", detail=detail,
+                         generation=generation)
+        return rec
+
+    def record_real_world(self, behaviour_id: str, *, passed: bool,
+                          detail: str = "", generation: str = "",
+                          confirmed_by: str = "user") -> dict:
+        """Real-world/manual test result (testing mode). USER_CONFIRMED
+        requires an explicit confirmed_by='user' pass (rule 19M)."""
+        rec = self.get(behaviour_id)
+        if rec is None:
+            raise ValueError(f"unknown behaviour {behaviour_id!r}")
+        rec["evidence"].append({"at": _now(), "kind": "real_world",
+                                "passed": passed,
+                                "detail": detail[:200],
+                                "generation": generation,
+                                "by": confirmed_by})
+        if passed:
+            state = (USER_CONFIRMED if confirmed_by == "user"
+                     else REAL_WORLD_PASSED)
+        else:
+            state = TEST_FAILED
+        self._transition(rec, state, kind="real_world", detail=detail,
+                         generation=generation)
+        return rec
+
+    def accept_limitation(self, behaviour_id: str, *, note: str,
+                          approved_by: str = "user") -> dict:
+        rec = self.get(behaviour_id)
+        if rec is None:
+            raise ValueError(f"unknown behaviour {behaviour_id!r}")
+        rec["limitations"].append({"at": _now(), "note": note[:200],
+                                   "approved_by": approved_by})
+        self._transition(rec, ACCEPTED_LIMITATION, kind="accept_limitation",
+                         detail=note, )
+        return rec
+
+    # -- selective invalidation (rule 19I) -----------------------------------
+
+    def invalidate_for_change(self, *, changed_entities: set[str],
+                              generation: str = "",
+                              reasons: dict[str, str] | None = None
+                              ) -> list[dict]:
+        """Mark RETEST_REQUIRED exactly those confirmed behaviours whose
+        structural dependencies intersect the changed closure. Every
+        invalidation carries a reason; unrelated confirmations stay.
+        Returns the invalidated records."""
+        reasons = reasons or {}
+        invalidated = []
+        for rec in self.doc["behaviours"].values():
+            if rec["state"] not in CONFIRMED_STATES:
+                continue
+            overlap = set(rec.get("dependencies") or []) & changed_entities
+            if rec.get("owner_component") in changed_entities:
+                overlap.add(rec["owner_component"])
+            if not overlap:
+                continue  # unrelated hardened behaviour: PRESERVED (19S-5)
+            because = reasons.get(
+                rec["behaviour_id"],
+                f"structural change touched: {sorted(overlap)}")
+            rec["invalidations"].append({
+                "at": _now(), "generation": generation,
+                "because": because[:200],
+                "was": rec["state"],
+                "changed_entities": sorted(overlap)})
+            # HISTORY IS PRESERVED — prior confirmation stays queryable
+            self._transition(rec, RETEST_REQUIRED,
+                             kind="selective_invalidation",
+                             detail=f"was {rec['state']}; {because}",
+                             generation=generation)
+            invalidated.append(rec)
+        return invalidated
+
+    def retest_history(self, behaviour_id: str) -> list[dict]:
+        """Generation-aware history: all confirmations + invalidations
+        for one behaviour, oldest first (rule 19S-12)."""
+        rec = self.get(behaviour_id)
+        if rec is None:
+            return []
+        events = []
+        for h in rec["history"]:
+            events.append({"at": h["at"], "generation": h.get("generation"),
+                           "event": f"{h['from']}→{h['to']}",
+                           "kind": h["kind"], "detail": h.get("detail", "")})
+        return events
+
+    # -- queue (testing mode, 19L/M) ------------------------------------------
+
+    def outstanding(self, *, component: str | None = None,
+                    project: str | None = None) -> list[dict]:
+        """Eligible testing items: anything not confirmed/accepted/n.a.
+        Dynamic — the queue re-derives from states after every fix
+        (rule 19M)."""
+        out = []
+        for rec in self.doc["behaviours"].values():
+            if rec["state"] in (USER_CONFIRMED, ACCEPTED_LIMITATION,
+                                NOT_APPLICABLE):
+                continue
+            if project and rec.get("owning_project") != project:
+                continue
+            if component and rec.get("owner_component") != component:
+                continue
+            out.append(rec)
+        out.sort(key=lambda r: (0 if r["state"] == TEST_FAILED else 1,
+                                r["behaviour_id"]))
+        return out
+
+    def summary(self) -> dict:
+        counts: dict[str, int] = {}
+        for rec in self.doc["behaviours"].values():
+            counts[rec["state"]] = counts.get(rec["state"], 0) + 1
+        return {"total": len(self.doc["behaviours"]),
+                "states": counts,
+                "outstanding": len(self.outstanding())}
@@ -0,0 +1,314 @@
+"""build_basis.py — T18 §18C: durable versioned RESEARCH BASIS for
+construction (V1 rule F: RESEARCH INFORMS STRUCTURE).
+
+Research output must not disappear in transient model context. A Build
+Basis is a durable, schema-versioned document that captures WHAT
+research decided and WHY, with references — not research dumps. It is
+created after the research-depth decision and before/with the
+Construction Contract, and is attachable to:
+
+  - a Project Scaffold (project-level basis)
+  - a Construction Contract (per-contract basis, via evidence_ref)
+  - future structural entities where useful
+
+Research Core ownership (rule C): this module NEVER researches. It
+records the depth decision (derived FROM adaptive_router's route — no
+second router), references the canonical pipeline's qualified output
+(stages ISSUE→…→QUALIFY, artifacts in research/issue-<N>/), and
+structures the adopt/adapt/combine/reject/innovate decisions (rule E).
+"""
+from __future__ import annotations
+
+import time
+from typing import Any
+
+SCHEMA_VERSION = 1
+
+# Research depth tiers — derived from adaptive_router ROUTES, never a
+# second router (rule C/A). Each tier records WHY it was chosen.
+DEPTH_NONE = "NONE"                # tiny obvious local change; local evidence only
+DEPTH_INWARD = "INWARD"            # repo-internal understanding suffices
+DEPTH_OUTWARD = "OUTWARD"          # precedent/prior-art discovery required
+DEPTH_FULL = "FULL"                # full pipeline incl. adversarial+synthesis
+
+DEPTH_TIERS = (DEPTH_NONE, DEPTH_INWARD, DEPTH_OUTWARD, DEPTH_FULL)
+
+# route → default depth (recorded as decision evidence; caller may
+# justify an exception, which is recorded as such — never silent)
+_ROUTE_DEFAULT_DEPTH = {
+    "LOCAL_FIX": DEPTH_NONE,
+    "STANDARD_CHANGE": DEPTH_INWARD,
+    "RESEARCHED_CHANGE": DEPTH_FULL,
+}
+
+
+def research_depth(route: str, *, mode: str = "",
+                   justification: str = "") -> dict:
+    """The research-depth DECISION for a construction need. Derived from
+    the canonical route authority (adaptive_router); `justification` is
+    REQUIRED when deviating from the route default — recorded, never
+    silent (rule D: record why research depth was chosen)."""
+    default = _ROUTE_DEFAULT_DEPTH.get(route, DEPTH_INWARD)
+    depth = default
+    why = f"route {route} default"
+    if justification and justification.strip():
+        # an explicit justification selects a tier by name if given;
+        # otherwise it annotates the default
+        candidate = justification.strip().split(":", 1)[0].strip()
+        if candidate in DEPTH_TIERS:
+            depth = candidate
+            why = justification.strip()
+        else:
+            why = f"route {route} default; note: {justification.strip()}"
+    elif route not in _ROUTE_DEFAULT_DEPTH:
+        why = f"route {route!r} unknown — INWARD floor (never silent NONE)"
+        depth = DEPTH_INWARD
+    return {"depth": depth, "why": why, "route": route, "mode": mode,
+            "decided_at": _now()}
+
+
+# OUTWARD precedent-first ordering contract (rule E): the questions the
+# prior-art pass MUST answer before deeper material is consulted.
+OUTWARD_PRIOR_ART_QUESTIONS = (
+    "what already exists addressing this intent",
+    "what open-source systems/products/workflows attempt it",
+    "what architectural patterns already work",
+    "what standards/conventions exist",
+    "what can be directly ADOPTED",
+    "what can be ADAPTED",
+    "what can be COMBINED",
+    "what should be REJECTED",
+    "where is genuine INNOVATION required",
+)
+
+_DISPOSITIONS = ("ADOPT", "ADAPT", "COMBINE", "REJECT", "INNOVATE")
+
+
+def new_basis(*, question: str, depth_decision: dict,
+              issue_ref: str = "") -> dict:
+    """Create an empty-but-valid Build Basis. References only — full
+    research dumps live in the Research Core's artifact dirs (rule 18C:
+    do not store unnecessary dumps)."""
+    return {
+        "schema_version": SCHEMA_VERSION,
+        "kind": "build_basis",
+        "created_at": _now(),
+        "question": question,
+        "issue_ref": issue_ref,  # e.g. research/issue-164 or Gitea issue
+        "depth": depth_decision,
+        "inward": {"findings": [], "refs": []},
+        "outward": {"prior_art_questions_answered": [],
+                    "findings": [], "refs": []},
+        "deeper": {"used": False, "findings": [], "refs": []},
+        "dispositions": [],  # [{item, disposition: ADOPT/…, why, ref}]
+        "alternatives_considered": [],
+        "uncertainty": [],
+        "risks": [],
+        "implications": {"architecture": [], "structural": [],
+                         "migration": [], "verification": []},
+    }
+
+
+def set_dispositions(basis: dict, dispositions: list[dict]) -> list[str]:
+    """Record the adopt/adapt/combine/reject/innovate decisions (rule E).
+    Returns validation errors; the basis stays unqualified until valid."""
+    basis["dispositions"] = [
+        {"item": d.get("item", ""), "disposition": d.get("disposition", ""),
+         "why": d.get("why", ""), "ref": d.get("ref", "")}
+        for d in dispositions
+    ]
+    return validate_basis(basis)
+
+
+def validate_basis(basis: dict) -> list[str]:
+    """Qualification gate. A basis is QUALIFIED when it carries a depth
+    decision with why, at least one outward prior-art answer when depth
+    is OUTWARD/FULL, and every disposition uses a legal disposition."""
+    errors: list[str] = []
+    if basis.get("schema_version") != SCHEMA_VERSION:
+        errors.append(f"schema_version {basis.get('schema_version')!r}")
+    depth = (basis.get("depth") or {}).get("depth")
+    if depth not in DEPTH_TIERS:
+        errors.append(f"depth {depth!r} not in {DEPTH_TIERS}")
+    if not (basis.get("depth") or {}).get("why"):
+        errors.append("depth decision lacks WHY (rule D)")
+    if depth in (DEPTH_OUTWARD, DEPTH_FULL):
+        outward = basis.get("outward") or {}
+        if not (outward.get("prior_art_questions_answered")
+                or outward.get("findings") or outward.get("refs")):
+            errors.append("OUTWARD/FULL depth requires prior-art evidence")
+    if depth == DEPTH_INWARD:
+        inward = basis.get("inward") or {}
+        if not (inward.get("findings") or inward.get("refs")
+                or basis.get("dispositions")):
+            errors.append("INWARD depth requires recorded evidence "
+                          "(findings/refs) — an empty basis is not "
+                          "qualified")
+    for d in basis.get("dispositions", []):
+        if d.get("disposition") not in _DISPOSITIONS:
+            errors.append(f"illegal disposition {d.get('disposition')!r} "
+                          f"for {d.get('item','')[:40]!r}")
+    return errors
+
+
+def is_qualified(basis: dict) -> bool:
+    return not validate_basis(basis)
+
+
+def attach_evidence_ref(basis: dict) -> str:
+    """The evidence_ref string a contract carries to point at this basis
+    (durable pointer, not a copy — stores stay independent)."""
+    return (f"build_basis:{basis.get('issue_ref') or 'adhoc'}:"
+            f"{basis.get('created_at')}")
+
+
+def summary(basis: dict) -> dict:
+    """Compact projection for prompts/previews (no private reasoning —
+    structured conclusions only, rule 18G)."""
+    return {
+        "schema_version": basis.get("schema_version"),
+        "question": basis.get("question"),
+        "issue_ref": basis.get("issue_ref"),
+        "depth": basis.get("depth"),
+        "qualified": is_qualified(basis),
+        "validation": validate_basis(basis),
+        "dispositions": [
+            {"item": d.get("item", "")[:80],
+             "disposition": d.get("disposition")}
+            for d in basis.get("dispositions", [])
+        ],
+        "uncertainty_count": len(basis.get("uncertainty", [])),
+        "risks_count": len(basis.get("risks", [])),
+        "implications": {k: len(v) for k, v in
+                         (basis.get("implications") or {}).items()},
+    }
+
+
+def _now() -> str:
+    return time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime())
+
+
+def revise_depth(basis: dict, *, new_depth: str, reason: str) -> dict:
+    """§21.1G: record a justified depth revision. Original + revised
+    depths both persist. An OUTWARD→INWARD/NONE downgrade requires
+    explicit resolution evidence ('discovery resolved', 'prior-art
+    resolved', 'unnecessary') — never a bare skip."""
+    if new_depth not in DEPTH_TIERS:
+        raise ValueError(f"unknown depth {new_depth!r}")
+    original = basis["depth"]["depth"]
+    if original == new_depth:
+        raise ValueError("no-op revision: depth unchanged")
+    consequential_origins = ("RESEARCHED_CHANGE", "FULL", "OUTWARD")
+    if original in consequential_origins and new_depth in ("NONE", "INWARD"):
+        keywords = ("resolved", "non-material", "discovery",
+                    "unnecessary", "prior-art question")
+        if not any(k in reason.lower() for k in keywords):
+            raise ValueError(
+                "OUTWARD downgrade requires explicit evidence: the "
+                "prior-art question must be resolved as non-material "
+                "(not merely skipped)")
+    basis["depth_revision"] = {
+        "original_depth": original, "revised_depth": new_depth,
+        "reason": reason[:200], "revised_at": _now()}
+    basis["depth"]["depth"] = new_depth
+    basis["depth"]["why"] = f"revised from {original}: {reason}"
+    return basis["depth_revision"]
+
+
+def to_json(basis: dict) -> str:
+    import json
+    return json.dumps(basis, indent=1, sort_keys=True)
+
+
+def from_json(text: str) -> dict:
+    import json
+    return json.loads(text)
+
+
+_ = Any  # typing import kept for forward annotations
+
+
+# ── T18 §18D CORRECTION (Nick, 2026-09-15): THE RESEARCH PLAN ITSELF IS
+# CONTEXTUALLY GENERATED. The prior-art categories researched for the
+# structural-model task (architecture graphs, monorepo tooling, etc.)
+# were CONTEXTUAL to that task — they must never become a fixed topic
+# list. For every request the plan is derived from:
+#   intent → target/current reality → material unknowns/decisions/risks
+#   → context-specific research topics → proportional depth.
+# There is NO permanent research-topic checklist anywhere in Scaffolding.
+
+def generate_research_plan(*, intent: str, target: str = "",
+                           current_reality: str = "",
+                           unknowns: list[str] | None = None,
+                           decisions: list[str] | None = None,
+                           risks: list[str] | None = None,
+                           depth: str | None = None) -> dict:
+    """Generate a CONTEXT-SPECIFIC research plan for one construction
+    intent. Topics are synthesized from the material unknowns/decisions/
+    risks of THIS intent — the function has no topic vocabulary and no
+    permanent category list. Depth, when not supplied, is chosen by
+    consequence heuristics recorded in the plan (why is always present)."""
+    unknowns = [u for u in (unknowns or []) if u.strip()]
+    decisions = [d for d in (decisions or []) if d.strip()]
+    risks = [r for r in (risks or []) if r.strip()]
+    topics: list[dict] = []
+    # each material unknown becomes a research topic aimed at resolving it
+    for u in unknowns:
+        topics.append({
+            "topic": f"resolve unknown: {u}",
+            "kind": "unknown-resolution",
+            "why": f"material unknown blocks confident construction: {u[:80]}",
+        })
+    for d in decisions:
+        topics.append({
+            "topic": f"inform decision: {d}",
+            "kind": "decision-support",
+            "why": f"consequential decision needs precedent/evidence: {d[:80]}",
+        })
+    for r in risks:
+        topics.append({
+            "topic": f"de-risk: {r}",
+            "kind": "risk-investigation",
+            "why": f"identified risk needs verification/mitigation evidence: {r[:80]}",
+        })
+    # precedent question (rule E) attaches ONLY when outward evidence is
+    # plausibly relevant — an unknown/decision about external prior art.
+    needs_outward = any(
+        w in " ".join(unknowns + decisions).lower()
+        for w in ("exist", "existing", "prior art", "precedent", "standard",
+                  "convention", "library", "tool", "framework", "approach"))
+    if needs_outward:
+        topics.append({
+            "topic": ("what already exists that is relevant to THIS intent "
+                      "(adopt/adapt/combine/reject/innovate)"),
+            "kind": "prior-art",
+            "why": "material unknown references existing practice — "
+                   "precedent discovery required before deeper research",
+        })
+    if depth is None:
+        # consequence heuristics: novel/architectural/unsafe → deeper
+        consequential = any(
+            w in (intent + " " + " ".join(risks)).lower()
+            for w in ("architecture", "migration", "replace", "rewrite",
+                      "security", "schema", "delete", "deploy", "breaking"))
+        depth = DEPTH_FULL if (consequential and topics) else (
+            DEPTH_OUTWARD if needs_outward and topics else (
+                DEPTH_INWARD if topics else DEPTH_NONE))
+        why = ("consequence heuristics: consequential keywords + research "
+               "topics present" if depth == DEPTH_FULL else
+               "precedent-relevant unknowns present" if depth == DEPTH_OUTWARD
+               else "local understanding suffices" if depth == DEPTH_INWARD
+               else "no material unknowns — local evidence only")
+    else:
+        why = "depth supplied by caller (recorded, caller accountable)"
+    return {
+        "intent": intent,
+        "target": target,
+        "current_reality_note": current_reality[:200],
+        "depth": {"depth": depth, "why": why},
+        "topics": topics,
+        "plan_generated_at": _now(),
+        # the plan ITSELF is evidence that research was contextual, not
+        # checklist-driven: two different intents must produce disjoint
+        # topic sets unless they share genuine unknowns
+    }
@@ -48,6 +48,7 @@ REPO_DIR = "/mnt/workdrive/rune"
 # DATA STRUCTURES
 # ═══════════════════════════════════════════════════════════════════════════════

+
 class Phase(Enum):
    ASSESS = "assess"
    RESEARCH = "research"
@@ -61,18 +62,23 @@ class Phase(Enum):
    INTEGRATE = "integrate"
    RECORD = "record"

+
 class BuildPhase(Enum):
    """Three-phase execution for build step."""
-    FOUNDATION = "foundation"      # Types, interfaces, shared utilities
+
+    FOUNDATION = "foundation"  # Types, interfaces, shared utilities
    IMPLEMENTATION = "implementation"  # Feature code
-    INTEGRATION = "integration"    # Tests, wiring, documentation
+    INTEGRATION = "integration"  # Tests, wiring, documentation
+

 class VerificationTier(Enum):
    """Tiered verification strategy."""
-    SYNTACTIC = "syntactic"    # Per-edit: lint, type-check, format (<1s)
-    SEMANTIC = "semantic"      # Per-batch: unit tests, import validation (<30s)
+
+    SYNTACTIC = "syntactic"  # Per-edit: lint, type-check, format (<1s)
+    SEMANTIC = "semantic"  # Per-batch: unit tests, import validation (<30s)
    INTEGRATION = "integration"  # Per-wave: contract tests, build/compile (<5min)
-    SYSTEM = "system"          # Full: full test suite, property-based (<15min)
+    SYSTEM = "system"  # Full: full test suite, property-based (<15min)
+

 class PhaseStatus(Enum):
    PENDING = "pending"
@@ -81,6 +87,7 @@ class PhaseStatus(Enum):
    FAILED = "failed"
    SKIPPED = "skipped"

+
@dataclass
 class LedgerEntry:
    phase: str
@@ -90,6 +97,7 @@ class LedgerEntry:
    details: Dict[str, Any] = field(default_factory=dict)
    errors: List[str] = field(default_factory=list)

+
@dataclass
 class BuildState:
    build_id: str
@@ -112,10 +120,12 @@ class BuildState:
    files_changed: List[str] = field(default_factory=list)
    contracts: List[Dict[str, Any]] = field(default_factory=list)

+
 # ═══════════════════════════════════════════════════════════════════════════════
 # BUILD MANAGER
 # ═══════════════════════════════════════════════════════════════════════════════

+
 class BuildManager:
    """Manages the build pipeline with worktree isolation and ledger tracking."""

@@ -282,11 +292,15 @@ class BuildManager:

    def complete_feasible(self, feasible: bool, risk: str, details: Dict[str, Any] = None):
        """Complete the feasibility phase."""
-        self._log_phase(Phase.FEASIBLE.value, PhaseStatus.COMPLETE.value, {
-            "feasible": feasible,
-            "risk": risk,
-            "details": details or {},
-        })
+        self._log_phase(
+            Phase.FEASIBLE.value,
+            PhaseStatus.COMPLETE.value,
+            {
+                "feasible": feasible,
+                "risk": risk,
+                "details": details or {},
+            },
+        )

    def start_consult(self):
        """Start the consult phase."""
@@ -327,10 +341,14 @@ class BuildManager:

    def complete_verify(self, passed: bool, failures: List[str] = None):
        """Complete the verify phase."""
-        self._log_phase(Phase.VERIFY.value, PhaseStatus.COMPLETE.value, {
-            "passed": passed,
-            "failures": failures or [],
-        })
+        self._log_phase(
+            Phase.VERIFY.value,
+            PhaseStatus.COMPLETE.value,
+            {
+                "passed": passed,
+                "failures": failures or [],
+            },
+        )

    def start_correct(self):
        """Start the correct phase."""
@@ -347,10 +365,14 @@ class BuildManager:

    def complete_integrate(self, passed: bool, failures: List[str] = None):
        """Complete the integrate phase."""
-        self._log_phase(Phase.INTEGRATE.value, PhaseStatus.COMPLETE.value, {
-            "passed": passed,
-            "failures": failures or [],
-        })
+        self._log_phase(
+            Phase.INTEGRATE.value,
+            PhaseStatus.COMPLETE.value,
+            {
+                "passed": passed,
+                "failures": failures or [],
+            },
+        )

    def start_record(self):
        """Start the record phase."""
@@ -409,6 +431,7 @@ class BuildManager:
 # CLI
 # ═══════════════════════════════════════════════════════════════════════════════

+
 def main():
    if len(sys.argv) < 2:
        print("Usage: build_mode.py [start|status|resume|abort]")
@@ -9,6 +9,7 @@ process is fresh each call, so the hook-level dedupe is the effective one).
 Usage: builder_bridge_cli.py "<task text>"
 Output: the capsule markdown (possibly empty — inject-nothing policy).
 """
+
 import json
 import os
 import sys
@@ -38,15 +38,15 @@ DB_PATH = os.path.join(memory_env.memory_dir(), "cache.db")
 # ---------------------------------------------------------------------------
 # In-memory hit tracking — avoids write-on-read in SQLite
 # ---------------------------------------------------------------------------
-_hit_counts: dict[str, int] = {}          # key → pending hit increments
-_last_hits: dict[str, str] = {}           # key → latest last_hit timestamp
-_miss_counts: dict[str, int] = {}         # key → pending miss increments
-_last_misses: dict[str, str] = {}         # key → latest last_miss timestamp
+_hit_counts: dict[str, int] = {}  # key → pending hit increments
+_last_hits: dict[str, str] = {}  # key → latest last_hit timestamp
+_miss_counts: dict[str, int] = {}  # key → pending miss increments
+_last_misses: dict[str, str] = {}  # key → latest last_miss timestamp
 _flush_lock = threading.Lock()
-_FLUSH_THRESHOLD = 50       # flush after this many pending updates
-_flush_counter = 0           # counts hits+misses since last flush
-_last_flush_time = 0.0       # epoch of last flush
-_FLUSH_INTERVAL = 30.0       # flush at least every N seconds
+_FLUSH_THRESHOLD = 50  # flush after this many pending updates
+_flush_counter = 0  # counts hits+misses since last flush
+_last_flush_time = 0.0  # epoch of last flush
+_FLUSH_INTERVAL = 30.0  # flush at least every N seconds


 def _record_hit(key: str) -> None:
@@ -58,8 +58,7 @@ def _record_hit(key: str) -> None:
        _hit_counts[key] = _hit_counts.get(key, 0) + 1
        _last_hits[key] = now
        _flush_counter += 1
-        if (_flush_counter >= _FLUSH_THRESHOLD
-                or (time.monotonic() - _last_flush_time) >= _FLUSH_INTERVAL):
+        if _flush_counter >= _FLUSH_THRESHOLD or (time.monotonic() - _last_flush_time) >= _FLUSH_INTERVAL:
            do_flush = True
    if do_flush:
        flush_stats()
@@ -74,8 +73,7 @@ def _record_miss(key: str) -> None:
        _miss_counts[key] = _miss_counts.get(key, 0) + 1
        _last_misses[key] = now
        _flush_counter += 1
-        if (_flush_counter >= _FLUSH_THRESHOLD
-                or (time.monotonic() - _last_flush_time) >= _FLUSH_INTERVAL):
+        if _flush_counter >= _FLUSH_THRESHOLD or (time.monotonic() - _last_flush_time) >= _FLUSH_INTERVAL:
            do_flush = True
    if do_flush:
        flush_stats()
@@ -187,10 +185,7 @@ def get_cache(key):
    """
    db = connect()
    try:
-        row = db.execute(
-            "SELECT * FROM cache WHERE key=? AND invalidated_at IS NULL",
-            (key,)
-        ).fetchone()
+        row = db.execute("SELECT * FROM cache WHERE key=? AND invalidated_at IS NULL", (key,)).fetchone()

        if row:
            # Record hit in memory — no SQLite write on the hot path
@@ -211,9 +206,7 @@ def get_cache(key):

        # Miss — only record if the key exists in the cache (e.g. invalidated).
        # A true miss (key never seen) should not trigger any write.
-        exists = db.execute(
-            "SELECT 1 FROM cache WHERE key=?", (key,)
-        ).fetchone()
+        exists = db.execute("SELECT 1 FROM cache WHERE key=?", (key,)).fetchone()

        if exists:
            _record_miss(key)
@@ -242,7 +235,8 @@ def put_cache(key, value, compressed=None, scope="general", source_hash=""):
    if compressed is None:
        compressed = _compress(value)

-    db.execute("""
+    db.execute(
+        """
        INSERT INTO cache
        (key, scope, content_hash, full_value, compressed_value, source_hash,
         created_at, invalidated_at, hit_count, miss_count)
@@ -255,7 +249,9 @@ def put_cache(key, value, compressed=None, scope="general", source_hash=""):
            source_hash = excluded.source_hash,
            created_at = excluded.created_at,
            invalidated_at = NULL
-    """, (key, scope, content_hash, value, compressed, source_hash, _now()))
+    """,
+        (key, scope, content_hash, value, compressed, source_hash, _now()),
+    )
    db.commit()
    db.close()
    return {"stored": key, "content_hash": content_hash, "scope": scope}
@@ -264,10 +260,7 @@ def put_cache(key, value, compressed=None, scope="general", source_hash=""):
 def invalidate(key, reason=""):
    """Invalidate a specific cache entry."""
    db = connect()
-    db.execute(
-        "UPDATE cache SET invalidated_at=? WHERE key=? AND invalidated_at IS NULL",
-        (_now(), key)
-    )
+    db.execute("UPDATE cache SET invalidated_at=? WHERE key=? AND invalidated_at IS NULL", (_now(), key))
    db.commit()
    db.close()
    return {"invalidated": key, "reason": reason}
@@ -277,8 +270,7 @@ def invalidate_scope(scope, reason=""):
    """Invalidate all entries in a scope."""
    db = connect()
    count = db.execute(
-        "UPDATE cache SET invalidated_at=? WHERE scope=? AND invalidated_at IS NULL",
-        (_now(), scope)
+        "UPDATE cache SET invalidated_at=? WHERE scope=? AND invalidated_at IS NULL", (_now(), scope)
    ).rowcount
    db.commit()
    db.close()
@@ -288,10 +280,7 @@ def invalidate_scope(scope, reason=""):
 def invalidate_all(reason=""):
    """Invalidate everything. Nuclear option."""
    db = connect()
-    count = db.execute(
-        "UPDATE cache SET invalidated_at=? WHERE invalidated_at IS NULL",
-        (_now(),)
-    ).rowcount
+    count = db.execute("UPDATE cache SET invalidated_at=? WHERE invalidated_at IS NULL", (_now(),)).rowcount
    db.commit()
    db.close()
    return {"invalidated_all": count, "reason": reason}
@@ -409,8 +398,7 @@ def handle_hook(event, data=None):

 def main():
    ap = argparse.ArgumentParser(description="Cache system — event-driven, write-once-derive-twice")
-    ap.add_argument("cmd", choices=["get", "put", "invalidate", "invalidate-all",
-                                     "invalidate-scope", "stats", "hook"])
+    ap.add_argument("cmd", choices=["get", "put", "invalidate", "invalidate-all", "invalidate-scope", "stats", "hook"])
    ap.add_argument("arg", nargs="?", default="")
    ap.add_argument("arg2", nargs="?", default="")
    ap.add_argument("--scope", default="general")
@@ -440,8 +428,11 @@ def main():
            print("usage: cache.py put <key> <value>")
            return
        result = put_cache(args.arg, args.arg2, scope=args.scope, source_hash=args.source_hash)
-        print(json.dumps(result, indent=2) if args.json else
-              f"Stored: {args.arg} (hash: {result['content_hash']}, scope: {result['scope']})")
+        print(
+            json.dumps(result, indent=2)
+            if args.json
+            else f"Stored: {args.arg} (hash: {result['content_hash']}, scope: {result['scope']})"
+        )

    elif args.cmd == "invalidate":
        if not args.arg:
@@ -452,16 +443,16 @@ def main():

    elif args.cmd == "invalidate-all":
        result = invalidate_all(reason=args.reason)
-        print(json.dumps(result, indent=2) if args.json else
-              f"Invalidated all: {result['invalidated_all']} entries")
+        print(json.dumps(result, indent=2) if args.json else f"Invalidated all: {result['invalidated_all']} entries")

    elif args.cmd == "invalidate-scope":
        if not args.arg:
            print("usage: cache.py invalidate-scope <scope>")
            return
        result = invalidate_scope(args.arg, reason=args.reason)
-        print(json.dumps(result, indent=2) if args.json else
-              f"Invalidated scope '{args.arg}': {result['count']} entries")
+        print(
+            json.dumps(result, indent=2) if args.json else f"Invalidated scope '{args.arg}': {result['count']} entries"
+        )

    elif args.cmd == "stats":
        s = cache_stats()
@@ -497,5 +488,5 @@ PLUGIN = {
    "dependencies": [],
    "provides": ["cache"],
    "priority": 1,
-    "auto_activate": True
+    "auto_activate": True,
 }
@@ -34,9 +34,7 @@ def main() -> int:
        runtime.registry.disable("synthesis")
        third = runtime.process({"event_type": "c"}, [{"event_type": "new"}])
        if {r["worker"] for r in third["results"]} != {"n1"}:
-            print(
-                "COGNITIVE CANARY FAILED: toggle independence broken", file=sys.stderr
-            )
+            print("COGNITIVE CANARY FAILED: toggle independence broken", file=sys.stderr)
            return 1
        print("COGNITIVE CANARY PASSED", third["facts"])
        return 0
@@ -41,6 +41,7 @@ def main():
    import audit_ledger
    import warm_router
    import memory_consolidation
+
    try:
        import curator
    except ImportError:
@@ -51,16 +52,20 @@ def main():

    # --- 1. Schema ---
    print("--- 1. Schema ---")
-    cols = {r[1] for r in db.execute('PRAGMA table_info(entries)')}
-    for c in ['wash_count', 'last_washed', 'promotion_eligible', 'rock_status']:
+    cols = {r[1] for r in db.execute("PRAGMA table_info(entries)")}
+    for c in ["wash_count", "last_washed", "promotion_eligible", "rock_status"]:
        test(f"column {c}", c in cols, f"{c} missing")

    # --- 2. Full lifecycle ---
    print("\n--- 2. Full lifecycle ---")
    # Find a washable entry (not already rock)
-    entry = db.execute('SELECT id, text, importance, wash_count, rock_status FROM entries WHERE status="active" AND rock_status="washable" LIMIT 1').fetchone()
+    entry = db.execute(
+        'SELECT id, text, importance, wash_count, rock_status FROM entries WHERE status="active" AND rock_status="washable" LIMIT 1'
+    ).fetchone()
    if not entry:
-        entry = db.execute('SELECT id, text, importance, wash_count, rock_status FROM entries WHERE status="active" LIMIT 1').fetchone()
+        entry = db.execute(
+            'SELECT id, text, importance, wash_count, rock_status FROM entries WHERE status="active" LIMIT 1'
+        ).fetchone()
    if not entry:
        test("full lifecycle", True, "no entries in database (skipped)")
        # Skip lifecycle tests, continue to other tests
@@ -73,8 +78,10 @@ def main():
        for i in range(20):
            memory_store.increment_wash(db, eid)

-        row = db.execute('SELECT id, text, importance, wash_count, rock_status FROM entries WHERE id=?', (eid,)).fetchone()
-        entry_dict = {'id': row[0], 'text': row[1], 'importance': row[2], 'wash_count': row[3], 'rock_status': row[4]}
+        row = db.execute(
+            "SELECT id, text, importance, wash_count, rock_status FROM entries WHERE id=?", (eid,)
+        ).fetchone()
+        entry_dict = {"id": row[0], "text": row[1], "importance": row[2], "wash_count": row[3], "rock_status": row[4]}

        # Value function
        value = drift_core.compute_value(entry_dict)
@@ -103,9 +110,11 @@ def main():
    print("\n--- 3. Cache ---")
    cache = drift_cache.DriftCache()
    if eid:
-        row = db.execute('SELECT id, text, importance, wash_count, rock_status FROM entries WHERE id=?', (eid,)).fetchone()
+        row = db.execute(
+            "SELECT id, text, importance, wash_count, rock_status FROM entries WHERE id=?", (eid,)
+        ).fetchone()
        if row:
-            cache.add(row[0], {'text': row[1], 'importance': row[2], 'wash_count': row[3], 'rock_status': row[4]})
+            cache.add(row[0], {"text": row[1], "importance": row[2], "wash_count": row[3], "rock_status": row[4]})
            test("cache contains", cache.contains(eid))
            test("cache priority", cache.get_priority(eid) > 0, f"priority={cache.get_priority(eid)}")
        else:
@@ -117,20 +126,29 @@ def main():

    # --- 4. Compression ---
    print("\n--- 4. Compression ---")
-    ratio_washable = drift_compression.get_compression_ratio(0, 'washable')
-    ratio_proven = drift_compression.get_compression_ratio(25, 'washable')
-    ratio_rock = drift_compression.get_compression_ratio(100, 'rock')
+    ratio_washable = drift_compression.get_compression_ratio(0, "washable")
+    ratio_proven = drift_compression.get_compression_ratio(25, "washable")
+    ratio_rock = drift_compression.get_compression_ratio(100, "rock")
    test("compression decreases with wash", ratio_washable > ratio_proven > ratio_rock)
-    test("should_warn for proven", drift_compression.should_warn({'wash_count': 15, 'rock_status': 'washable', 'text': 'x' * 300}))
-    test("warn_before_compress returns dict", isinstance(drift_compression.warn_before_compress({'wash_count': 15, 'rock_status': 'washable', 'text': 'x' * 300}), dict))
+    test(
+        "should_warn for proven",
+        drift_compression.should_warn({"wash_count": 15, "rock_status": "washable", "text": "x" * 300}),
+    )
+    test(
+        "warn_before_compress returns dict",
+        isinstance(
+            drift_compression.warn_before_compress({"wash_count": 15, "rock_status": "washable", "text": "x" * 300}),
+            dict,
+        ),
+    )

    # --- 5. Warm router ---
    print("\n--- 5. Warm router ---")
    neurons = warm_router.list_neurons()
    if neurons:
-        test("neurons have wash_count", any('wash_count' in n for n in neurons))
-        test("neurons have body_knowledge", any('body_knowledge' in n for n in neurons))
-        test("body_knowledge populated", sum(1 for n in neurons if n.get('body_knowledge', '')) > 0)
+        test("neurons have wash_count", any("wash_count" in n for n in neurons))
+        test("neurons have body_knowledge", any("body_knowledge" in n for n in neurons))
+        test("body_knowledge populated", sum(1 for n in neurons if n.get("body_knowledge", "")) > 0)
    else:
        test("neurons have wash_count", True, "no neurons in database (skipped)")
        test("neurons have body_knowledge", True, "no neurons in database (skipped)")
@@ -141,19 +159,20 @@ def main():
    mc = memory_consolidation.MemoryConsolidation()
    # Check if ConsolidatedKnowledge has wash_count field
    import dataclasses
+
    try:
        fields = [f.name for f in dataclasses.fields(memory_consolidation.ConsolidatedKnowledge)]
-        test("consolidation has wash_count", 'wash_count' in fields)
+        test("consolidation has wash_count", "wash_count" in fields)
    except Exception:
        test("consolidation has wash_count", False, "could not inspect dataclass")
    status = mc.get_status()
-    test("consolidation has entries", status['knowledge_count'] > 0 or "no entries in database")
+    test("consolidation has entries", status["knowledge_count"] > 0 or "no entries in database")

    # --- 7. Curator ---
    print("\n--- 7. Curator ---")
    if curator:
-        result = curator.le_promote_eligible('test', ['test'], 2, wash_count=15)
-        test("curator adjusted threshold", result['eligible'])
+        result = curator.le_promote_eligible("test", ["test"], 2, wash_count=15)
+        test("curator adjusted threshold", result["eligible"])
    else:
        test("curator adjusted threshold", True, "curator not available in TUI (skipped)")

@@ -162,20 +181,20 @@ def main():
    obs = drift_observability.DriftObservability()
    vel = drift_core.compute_drift_velocity(db)
    direction = drift_core.compute_drift_direction(db)
-    test("velocity computed", 'changes_per_day' in vel)
-    test("direction computed", 'stability_ratio' in direction)
+    test("velocity computed", "changes_per_day" in vel)
+    test("direction computed", "stability_ratio" in direction)

    # --- 9. Audit ledger ---
    print("\n--- 9. Audit ledger ---")
    chain = audit_ledger.verify_chain()
-    test("chain valid", chain['valid'], chain.get('reason', ''))
+    test("chain valid", chain["valid"], chain.get("reason", ""))

    # --- 10. Subconscious ---
    print("\n--- 10. Subconscious ---")
    sp = subconscious_pass.SubconsciousPass()
    report = sp.reflect("I am frustrated", "I understand")
-    test("reflection works", report['finding_count'] >= 0)
-    test("emotional detection", any(f['type'] == 'emotional_signal' for f in report['findings']))
+    test("reflection works", report["finding_count"] >= 0)
+    test("emotional detection", any(f["type"] == "emotional_signal" for f in report["findings"]))

    # --- 11. Retrieval with drift ---
    print("\n--- 11. Retrieval with drift ---")
@@ -40,24 +40,14 @@ def main() -> int:
    backup_script = (ROOT / "scripts" / "gitea-backup.sh").read_text()
    restore_script = (ROOT / "scripts" / "gitea-restore-verify.sh").read_text()
    rune_version = (ROOT / "src" / "rune" / "__init__.py").read_text()
-    gitea_sources = "\n".join(
-        path.read_text() for path in (ROOT / "shared").glob("gitea*.py")
-    )
+    gitea_sources = "\n".join(path.read_text() for path in (ROOT / "shared").glob("gitea*.py"))
    gitea_sources += (ROOT / "shared" / "custodian.py").read_text()

    print("=== GITEA AUTOMATION CANARY ===")
-    check(
-        "feature issue template exists", (issue_dir / "feature.md").is_file(), failures
-    )
+    check("feature issue template exists", (issue_dir / "feature.md").is_file(), failures)
    check("bug issue template exists", (issue_dir / "bug.md").is_file(), failures)
-    config = (
-        (issue_dir / "config.yml").read_text()
-        if (issue_dir / "config.yml").exists()
-        else ""
-    )
-    check(
-        "blank issues are disabled", "blank_issues_enabled: false" in config, failures
-    )
+    config = (issue_dir / "config.yml").read_text() if (issue_dir / "config.yml").exists() else ""
+    check("blank issues are disabled", "blank_issues_enabled: false" in config, failures)
    check("documented PR template path exists", pr_template.is_file(), failures)
    check(
        "PR template requires CI gate",
@@ -80,8 +70,7 @@ def main() -> int:
    )
    check(
        "auto-label declares every managed label",
-        "managedLabels" in auto_label
-        and all(f"'{label}'" in auto_label for label in managed_labels),
+        "managedLabels" in auto_label and all(f"'{label}'" in auto_label for label in managed_labels),
        failures,
    )
    check(
@@ -91,9 +80,7 @@ def main() -> int:
    )
    check(
        "stale managed labels are removed",
-        "managedLabels.has" in auto_label
-        and "!desiredLabels.has" in auto_label
-        and "removeLabel" in auto_label,
+        "managedLabels.has" in auto_label and "!desiredLabels.has" in auto_label and "removeLabel" in auto_label,
        failures,
    )
    check(
@@ -103,9 +90,7 @@ def main() -> int:
    )
    check(
        "race contract rejects inverted 404 handling",
-        not has_idempotent_label_removal(
-            auto_label.replace("error.status !== 404", "error.status === 404", 1)
-        ),
+        not has_idempotent_label_removal(auto_label.replace("error.status !== 404", "error.status === 404", 1)),
        failures,
    )
    check(
@@ -130,8 +115,7 @@ def main() -> int:
    )
    check(
        "backup is stored off the Pi root disk",
-        "/media/" in backup_script
-        and "/tmp/rune-backups" not in backup + backup_script,
+        "/media/" in backup_script and "/tmp/rune-backups" not in backup + backup_script,
        failures,
    )
    check(
@@ -158,8 +142,7 @@ def main() -> int:
    )
    check(
        "backup atomically publishes without clobber and links checksum last",
-        backup_script.rfind('ln -T -- "$STAGED_CHECKSUM"')
-        > backup_script.rfind('ln -T -- "$STAGED_ARCHIVE"')
+        backup_script.rfind('ln -T -- "$STAGED_CHECKSUM"') > backup_script.rfind('ln -T -- "$STAGED_ARCHIVE"')
        and 'mv -f -- "$STAGED_' not in backup_script
        and '"$BACKUP_DIR/$ARCHIVE.sha256" -ef "$STAGED_CHECKSUM"' in backup_script
        and "Removing the staged checksum disarms rollback" in backup_script,
@@ -173,14 +156,12 @@ def main() -> int:
    )
    check(
        "backup workflow pins audited checkout v4.2.2 commit",
-        "# actions/checkout v4.2.2" in backup
-        and "actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683" in backup,
+        "# actions/checkout v4.2.2" in backup and "actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683" in backup,
        failures,
    )
    check(
        "restore verifier is explicitly guarded and rootless",
-        "RUNE_GITEA_RESTORE_VERIFY" in restore_script
-        and "EUID != 0" in restore_script,
+        "RUNE_GITEA_RESTORE_VERIFY" in restore_script and "EUID != 0" in restore_script,
        failures,
    )
    check(
@@ -197,9 +178,7 @@ def main() -> int:
    )
    check(
        "restore SQLite import denies host-file attachment semantically",
-        "set_authorizer" in restore_script
-        and "SQLITE_ATTACH" in restore_script
-        and "SQLITE_DETACH" in restore_script,
+        "set_authorizer" in restore_script and "SQLITE_ATTACH" in restore_script and "SQLITE_DETACH" in restore_script,
        failures,
    )
    check(
@@ -221,7 +200,7 @@ def main() -> int:
        'cp -a -- "$RESTORE_ROOT/data/." "$INSTALL_ROOT/gitea/"' in restore_script
        and restore_script.index('cp -a -- "$RESTORE_ROOT/data/."')
        < restore_script.index('cp "$DB" "$INSTALL_ROOT/gitea/gitea.db"')
-        < restore_script.index("cat >\"$INSTALL_ROOT/gitea/conf/app.ini\""),
+        < restore_script.index('cat >"$INSTALL_ROOT/gitea/conf/app.ini"'),
        failures,
    )
    check(
@@ -252,7 +231,7 @@ def main() -> int:
    )
    check(
        "Rune minor version advertises provisioning and restore verification",
-        '__version__ = "0.66.4"' in rune_version,
+        '__version__ = "0.86.0"' in rune_version,
        failures,
    )
    # #169 policy update: auto-bump on merge is sanctioned (Nick directive:
@@ -15,17 +15,13 @@ from rune.orchestration.transport import FakeGiteaTransport
 def main() -> int:
    with tempfile.TemporaryDirectory() as td:
        auth = build_gitea_delegation(Path(td))
-        approval = ApprovalRecord(
-            approver="nick", plan_version="p", approved_items=("g1", "g2")
-        )
+        approval = ApprovalRecord(approver="nick", plan_version="p", approved_items=("g1", "g2"))
        records = auth.delegate(approval, ["g1", "g2"], FakeGiteaTransport())
        if len(records) != 3 or any(r.state != "accepted" for r in records):
            print("GITEA DELEGATION CANARY FAILED: round trip", file=sys.stderr)
            return 1
        if auth.store.pending() != []:
-            print(
-                "GITEA DELEGATION CANARY FAILED: pending not drained", file=sys.stderr
-            )
+            print("GITEA DELEGATION CANARY FAILED: pending not drained", file=sys.stderr)
            return 1
        print("GITEA DELEGATION CANARY PASSED", len(records))
        return 0
@@ -25,9 +25,7 @@ def main() -> int:
    pull_request_section = text.split("  pull_request:", 1)[1].split("\njobs:", 1)[0]
    required_gate = text.split("  required-checks:", 1)[1]
    offline_job = (
-        text.split("  gitea-offline:", 1)[1].split("  required-checks:", 1)[0]
-        if "  gitea-offline:" in text
-        else ""
+        text.split("  gitea-offline:", 1)[1].split("  required-checks:", 1)[0] if "  gitea-offline:" in text else ""
    )

    print("=== GITEA PIPELINE CANARY ===")
@@ -57,8 +55,7 @@ def main() -> int:
    check("isolates stateful canaries from host data", "RUNE_HOME:" in text, failures)
    check(
        "runs full test suite via auto-discovery",
-        "unit-tests:" in text
-        and "pytest tests/" in text,
+        "unit-tests:" in text and "pytest tests/" in text,
        failures,
    )
    # Full-suite auto-discovery: no explicit paths to validate.
@@ -96,6 +93,31 @@ def main() -> int:
        "if: always()" in required_gate and "needs: [lint, unit-tests, canary" in required_gate,
        failures,
    )
+    # RELEASE-SAFETY (2026-09-13, after 0.66.28/29 red promotions): the
+    # required-checks publish step previously echoed success UNCONDITIONALLY
+    # (read /tmp/need_* files that no step wrote). The verdict MUST come
+    # from the needs context — this canary locks that class shut.
+    publish_section = required_gate.split("Publish protected-branch gate", 1)[-1]
+    check(
+        "required gate verdict evaluates needs results (not unconditional echo)",
+        "needs.lint.result == 'success'" in publish_section
+        and "needs['type-check'].result == 'success'" in publish_section
+        and "needs['unit-tests'].result == 'success'" in publish_section,
+        failures,
+    )
+    check(
+        "required gate fails closed when any check is not success",
+        "!= 'success'" in publish_section and "exit 1" in publish_section,
+        failures,
+    )
+    check(
+        "no unconditional success echo without a needs-conditioned if",
+        not (
+            "All required CI checks passed" in publish_section
+            and "needs.lint.result == 'success'" not in publish_section
+        ),
+        failures,
+    )
    check(
        "required gate depends on unit tests and verification jobs",
        "needs: [lint, unit-tests, canary" in text,
@@ -113,8 +135,7 @@ def main() -> int:
    )
    check(
        "offline Gitea job syntax-checks both shell scripts",
-        "bash -n scripts/gitea-backup.sh scripts/gitea-restore-verify.sh"
-        in offline_job,
+        "bash -n scripts/gitea-backup.sh scripts/gitea-restore-verify.sh" in offline_job,
        failures,
    )
    check(
@@ -125,9 +146,7 @@ def main() -> int:
    check(
        "CI checkout actions use the audited v4.2.2 commit",
        "actions/checkout@v4" not in text
-        and text.count(
-            "actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683"
-        )
+        and text.count("actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683")
        == text.count("uses: actions/checkout@"),
        failures,
    )
@@ -14,12 +14,8 @@ from rune.orchestration.memory_synthesis import build_memory_synthesis
 def main() -> int:
    with tempfile.TemporaryDirectory() as td:
        auth = build_memory_synthesis(Path(td))
-        e1 = auth.episodic.append(
-            EpisodicRecord("", "user.message", ("sha256:a",), "t0")
-        )
-        e2 = auth.episodic.append(
-            EpisodicRecord("", "user.message", ("sha256:b",), "t1")
-        )
+        e1 = auth.episodic.append(EpisodicRecord("", "user.message", ("sha256:a",), "t0"))
+        e2 = auth.episodic.append(EpisodicRecord("", "user.message", ("sha256:b",), "t1"))
        r1 = auth.synthesize([e1.episode_id, e2.episode_id], "sha256:summary1")
        r2 = auth.synthesize([e1.episode_id, e2.episode_id], "sha256:summary2")
        if r1.covered != 2 or r2.covered != 0:
@@ -27,9 +23,7 @@ def main() -> int:
            return 1
        anc = [n.summary_ref for n in auth.chain.ancestors()]
        if anc != ["sha256:summary2", "sha256:summary1"]:
-            print(
-                "MEMORY SYNTHESIS CANARY FAILED: chain not cumulative", file=sys.stderr
-            )
+            print("MEMORY SYNTHESIS CANARY FAILED: chain not cumulative", file=sys.stderr)
            return 1
        print("MEMORY SYNTHESIS CANARY PASSED", len(anc))
        return 0