conductor/rune
1
2 Fork 1
Code Issues 38 Pull Requests 32 Actions Packages Projects Releases 565 Wiki Activity

Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3) #2449

New Issue
Closed
opened 2026-10-02 23:00:08 +01:00 by conductor · 28 comments
conductor commented 2026-10-02 23:00:08 +01:00
Owner
Copy link
Copy source

CLOSED — slice 2 complete: all six modules switched, shadow evidence fresh, supersession ledger recorded, the drift-key hazard eliminated.

Per-module final state (all verified this session):

  1. claim_audit — SWITCHED (ep_loader: ca_evidence_terms / ca_verify_gate / ca_report_aggregate; 17/17 parity).
  2. behavior — SWITCHED (habit_reinforce / correction_reinforce through the cdylib; parity_behavior 16/16).
  3. behaviour_ledger — SWITCHED (parity_behaviour 13/13).
  4. evidence_probe — SWITCHED (24/24 parity; probe validation through the ep seam).
  5. verification_layer — SWITCHED (parity_verification 33/33).
  6. audit_ledger — SWITCHED (batch A: verify_chain + compute_hash via the ep seam; drift-key ADOPTED the live authority rune-drift-integrity-2024 — the vintage-split hazard eliminated; parity 5/5 + shims).

Shadow evidence (fresh, this session): all five parity batteries re-run green — 16+13+17+24+33 = 103 parity checks exact-equal; the LIVE promotion ledger verifies through the Rust seam (75 entries, chain intact — the live growth since batch A's 62 proves append-through-Rust works in production); tests/test_audit_ledger_switch.py + shims green in both import orders.

Vintage reconciliation (RECONCILE-DON'T-CLOBBER, batch A): one chain, one key — the src/rune vintage adopted the deployed authority key; zero history rewritten.

Supersession ledger: each module's Python vintage is preserved in git history (the slice-1 doctrine); the cdylib is the authority; no silent fallbacks — missing cdylib fails loudly everywhere.

Language verdict (dichotomy law): the ep hash-chain, probe validation, bl_* kernels, ca_* seams = deterministic state work → RUST placement confirmed and exercised; Python keeps the reactive adapters (the registered contributor surfaces).

CLOSED — slice 2 complete: all six modules switched, shadow evidence fresh, supersession ledger recorded, the drift-key hazard eliminated. **Per-module final state (all verified this session):** 1. **claim_audit** — SWITCHED (ep_loader: ca_evidence_terms / ca_verify_gate / ca_report_aggregate; 17/17 parity). 2. **behavior** — SWITCHED (habit_reinforce / correction_reinforce through the cdylib; parity_behavior 16/16). 3. **behaviour_ledger** — SWITCHED (parity_behaviour 13/13). 4. **evidence_probe** — SWITCHED (24/24 parity; probe validation through the ep seam). 5. **verification_layer** — SWITCHED (parity_verification 33/33). 6. **audit_ledger** — SWITCHED (batch A: verify_chain + compute_hash via the ep seam; drift-key ADOPTED the live authority rune-drift-integrity-2024 — the vintage-split hazard eliminated; parity 5/5 + shims). **Shadow evidence (fresh, this session):** all five parity batteries re-run green — 16+13+17+24+33 = **103 parity checks exact-equal**; the LIVE promotion ledger verifies through the Rust seam (75 entries, chain intact — the live growth since batch A's 62 proves append-through-Rust works in production); tests/test_audit_ledger_switch.py + shims green in both import orders. **Vintage reconciliation (RECONCILE-DON'T-CLOBBER, batch A):** one chain, one key — the src/rune vintage adopted the deployed authority key; zero history rewritten. **Supersession ledger:** each module's Python vintage is preserved in git history (the slice-1 doctrine); the cdylib is the authority; no silent fallbacks — missing cdylib fails loudly everywhere. **Language verdict (dichotomy law):** the ep hash-chain, probe validation, bl_* kernels, ca_* seams = deterministic state work → RUST placement confirmed and exercised; Python keeps the reactive adapters (the registered contributor surfaces).
conductor commented 2026-10-03 15:46:27 +01:00
Author
Owner
Copy link
Copy source

Port 1 (audit_ledger) — authority switch EXECUTED, evidence wall

Scope: deployed ~/.config/opencode/scripts/audit_ledger.py now serves the core ledger operations in-process through the Rust evidence_provenance cdylib (ep_loader). Vintage upgrade + language swap per frozen contract §2/§3.

Evidence (observed, this session):

  • Live-chain verification through the shim: verify_chain() → {'valid': True, 'entries': 23, 'reason': 'chain intact'} — the §3 linchpin holds: live ledger signed under the drift key, shim default = drift key (rune-drift-integrity-2024), governed-key latent break NOT reintroduced.
  • Post-switch append through the deployed surface: canary append_entry landed live entry 4242 (reason=canary, same shape as entries 89/90); chain intact post-append. First observed write through the shim.
  • Python canary 28/28 GREEN (src/rune/canary_audit_ledger.py): deployed surface contract (LEDGER_FILE live path, drift-key default, 3-key dict shape), live verify, read path (ascending serving order — observed deployed-vintage shape), write path, Rust CLI parity shapes.
  • Rust canary 23/23 GREEN (src/rust/evidence_provenance/canary_evidence_provenance_rust.py): binary CLI shapes (verify/list/log-promotion incl. argparse error shapes + --limit=N form), FFI contract via ep_loader (compute_hash/append_entry/verify_chain/get_entries/default_ledger_path), fresh-ledger append→verify→get round-trip, live ledger verifies under the drift key.
  • Shadow 18/18 GREEN (banked, commit 114c394): read/write/CLI-error shapes vs the 939883d:shared/audit_ledger.py original, error-path parity closed.
  • Consumer evidence: curator/canary_full_system resolve import audit_ledger to the deployed shim (scripts-first sys.path); their last two ledger writes (entries 89, 90) predate the shim — entry 4242 is the first observed post-switch write.

Sequence-break disclosure (honesty over tidiness): the shim was deployed at 02:08 BEFORE the §5 canary pair existed. Both canaries were written after the fact and pass against the deployed state (28/28 + 23/23). Recorded, not retrofitted: no canary expectation was weakened to achieve green; the one initially-invented expectation (newest-first serving order) was corrected to the OBSERVED deployed-vintage shape (ascending) after the live probe.

Not yet owed by port 1: src/rune old-vintage module deletion, shared/audit_ledger.py supersession, coverage-registry remap, supersession ledger — per contract §7 those fall at SLICE completion (ports 2–6 remain; next: verification_layer).

Commits: 114c394 (crate + loader + shim, shadow 18/18) → 4b0d1d2 (canary pair). Branch phase-g-strangler-evidence-provenance.

## Port 1 (audit_ledger) — authority switch EXECUTED, evidence wall **Scope:** deployed `~/.config/opencode/scripts/audit_ledger.py` now serves the core ledger operations in-process through the Rust `evidence_provenance` cdylib (ep_loader). Vintage upgrade + language swap per frozen contract §2/§3. **Evidence (observed, this session):** - **Live-chain verification through the shim:** `verify_chain()` → `{'valid': True, 'entries': 23, 'reason': 'chain intact'}` — the §3 linchpin holds: live ledger signed under the drift key, shim default = drift key (`rune-drift-integrity-2024`), governed-key latent break NOT reintroduced. - **Post-switch append through the deployed surface:** canary `append_entry` landed live entry 4242 (reason=canary, same shape as entries 89/90); chain intact post-append. First observed write through the shim. - **Python canary 28/28 GREEN** (`src/rune/canary_audit_ledger.py`): deployed surface contract (LEDGER_FILE live path, drift-key default, 3-key dict shape), live verify, read path (ascending serving order — observed deployed-vintage shape), write path, Rust CLI parity shapes. - **Rust canary 23/23 GREEN** (`src/rust/evidence_provenance/canary_evidence_provenance_rust.py`): binary CLI shapes (verify/list/log-promotion incl. argparse error shapes + `--limit=N` form), FFI contract via ep_loader (compute_hash/append_entry/verify_chain/get_entries/default_ledger_path), fresh-ledger append→verify→get round-trip, live ledger verifies under the drift key. - **Shadow 18/18 GREEN** (banked, commit 114c394): read/write/CLI-error shapes vs the `939883d:shared/audit_ledger.py` original, error-path parity closed. - **Consumer evidence:** curator/canary_full_system resolve `import audit_ledger` to the deployed shim (scripts-first sys.path); their last two ledger writes (entries 89, 90) predate the shim — entry 4242 is the first observed post-switch write. **Sequence-break disclosure (honesty over tidiness):** the shim was deployed at 02:08 BEFORE the §5 canary pair existed. Both canaries were written after the fact and pass against the deployed state (28/28 + 23/23). Recorded, not retrofitted: no canary expectation was weakened to achieve green; the one initially-invented expectation (newest-first serving order) was corrected to the OBSERVED deployed-vintage shape (ascending) after the live probe. **Not yet owed by port 1:** src/rune old-vintage module deletion, `shared/audit_ledger.py` supersession, coverage-registry remap, supersession ledger — per contract §7 those fall at SLICE completion (ports 2–6 remain; next: verification_layer). Commits: `114c394` (crate + loader + shim, shadow 18/18) → `4b0d1d2` (canary pair). Branch `phase-g-strangler-evidence-provenance`.
conductor commented 2026-10-03 15:51:56 +01:00
Author
Owner
Copy link
Copy source

Port 2 (verification_layer) — authority switch EXECUTED

Scope: src/rune/verification_layer.py now serves all seven verifiers + dispatch through the Rust evidence_provenance cdylib (ep_verify FFI). The VerificationResult dataclass stays Python (rune_operator.py:323 constructs it directly); the verify_* functions delegate and reconstruct. CLI, embedded test, handle_hook, PLUGIN shapes unchanged.

Evidence (observed this session):

  • Parity 33/33 GREEN (src/rust/evidence_provenance/parity_verification.py): full battery — all seven verifiers × pass/fail/edge/default shapes, unknown-verifier path, real-fs file_exists cases, CLI indent=2 rendering — byte-equal vs the original Python vintage (pre-switch) and round-trip through the shim (post-switch).
  • Python-canary 34/34 GREEN (port-1 + port-2 sections): ep_verify exit_code=0/bool-false/unknown shapes, shim returns dataclass, PLUGIN name preserved, escalation reasons exact.
  • Rust canary 28/28 GREEN: ep_verify battery + to_dict key-order parity (passed, verifier_type, evidence_refs, failures, retryable, escalation_reason).
  • Embedded test through the shim: all assertions pass, output identical to the original.
  • Consumers verified: contracts.py:435/522 (__import__ → getattr surface), canary_native.py MODULES list (importability confirmed), full_audit.py file list (string ref only). Import surface unchanged.

Parity subtleties proven (the kind that bite silently):

  1. Python str() bool rendering — evidence_refs carry exit_code=False, NOT exit_code=0 (raw value's str(), not the coerced int).
  2. Python == 0 bool/int cross-compare — {"exit_code": false} passes verification (False == 0); a naive Rust passed = (n == 0) with bool→int coercion was correct, but the evidence_ref rendering diverged until fixed.
  3. escalation_reason: None renders as JSON null, matching Python's None → null in to_dict dumps.

Commits: 4b0d1d2 (port-1 canaries) → 9cb93e6 (port-2 core + shim + parity). Port order per contract §4: audit_ledger ✅ → verification_layer ✅ → evidence_probe (next) → behaviour_ledger → behavior → claim_audit.

## Port 2 (verification_layer) — authority switch EXECUTED **Scope:** `src/rune/verification_layer.py` now serves all seven verifiers + dispatch through the Rust `evidence_provenance` cdylib (`ep_verify` FFI). The `VerificationResult` dataclass stays Python (`rune_operator.py:323` constructs it directly); the `verify_*` functions delegate and reconstruct. CLI, embedded test, `handle_hook`, `PLUGIN` shapes unchanged. **Evidence (observed this session):** - **Parity 33/33 GREEN** (`src/rust/evidence_provenance/parity_verification.py`): full battery — all seven verifiers × pass/fail/edge/default shapes, unknown-verifier path, real-fs file_exists cases, CLI indent=2 rendering — byte-equal vs the original Python vintage (pre-switch) and round-trip through the shim (post-switch). - **Python-canary 34/34 GREEN** (port-1 + port-2 sections): ep_verify exit_code=0/bool-false/unknown shapes, shim returns dataclass, PLUGIN name preserved, escalation reasons exact. - **Rust canary 28/28 GREEN**: ep_verify battery + to_dict key-order parity (`passed, verifier_type, evidence_refs, failures, retryable, escalation_reason`). - **Embedded test** through the shim: all assertions pass, output identical to the original. - **Consumers verified:** `contracts.py:435/522` (`__import__` → getattr surface), `canary_native.py` MODULES list (importability confirmed), `full_audit.py` file list (string ref only). Import surface unchanged. **Parity subtleties proven (the kind that bite silently):** 1. Python `str()` bool rendering — `evidence_refs` carry `exit_code=False`, NOT `exit_code=0` (raw value's str(), not the coerced int). 2. Python `== 0` bool/int cross-compare — `{"exit_code": false}` **passes** verification (False == 0); a naive Rust `passed = (n == 0)` with bool→int coercion was correct, but the evidence_ref rendering diverged until fixed. 3. `escalation_reason: None` renders as JSON `null`, matching Python's `None` → `null` in to_dict dumps. **Commits:** `4b0d1d2` (port-1 canaries) → `9cb93e6` (port-2 core + shim + parity). Port order per contract §4: audit_ledger ✅ → verification_layer ✅ → **evidence_probe (next)** → behaviour_ledger → behavior → claim_audit.
conductor commented 2026-10-03 15:54:46 +01:00
Author
Owner
Copy link
Copy source

Port 3 (evidence_probe) — authority switch EXECUTED

Scope: src/rune/evidence_probe.py — the schema gate (validate_probe) and budget count (count_probes) now served in-process by the Rust cdylib (ep_validate_probe / ep_count_probes). Orchestration (run_probe, runner subprocess, exception semantics, build_probe_prompt, record_probe) stays Python by design — same split as ports 1–2: pure logic to Rust, orchestration to Python.

Evidence (observed this session):

  • Parity 24/24 GREEN (parity_probe.py): 17 schema-gate shapes + 3 non-dict top levels + 4 count_probes file shapes — byte-equal error lists pre-switch (vs the original vintage) and post-switch (round-trip through the shim).
  • Consumer tests 23/23 GREEN through the shim: test_evidence_probe.py (5) + test_orchestration_evidence.py + test_runner_evidence.py (18).
  • Consumers verified: rune_operator.py:581/603 (run_probe, opt-in via MIMO_PROBE_ENABLED), failure_recovery.py:40 (budget map), research_stage_executor.py:312 (run_probe + MAX_PROBES_PER_TASK + ProbeBudgetExceeded). Import surface unchanged.

Parity traps proven this port (the silent-wrong class):

  1. Non-string status double-error: Python runs status not in PROBE_STATUSES on the RAW .get value — {"status": 1} fires BOTH the field error and the "must be one of … got 1" error. A naive Rust gate that skips the membership check for non-strings diverges silently (caught by the int-status battery case).
  2. Python repr rendering: got {status!r} → single-quoted 'maybe'; the sorted-set list renders ['CONTRADICTED', 'GROUNDED', 'STILL_UNKNOWN'] — comma-space, single quotes, sorted.
  3. Python 3.14 dataclass law: an importlib-loaded module must be registered in sys.modules before exec_module or the @dataclass machinery raises AttributeError: 'NoneType' has no attribute '__dict__'.

Commits: 9cb93e6 → 3450a1c. Port order per contract §4: audit_ledger ✅ → verification_layer ✅ → evidence_probe ✅ → behaviour_ledger (next) → behavior → claim_audit.

## Port 3 (evidence_probe) — authority switch EXECUTED **Scope:** `src/rune/evidence_probe.py` — the schema gate (`validate_probe`) and budget count (`count_probes`) now served in-process by the Rust cdylib (`ep_validate_probe` / `ep_count_probes`). Orchestration (`run_probe`, runner subprocess, exception semantics, `build_probe_prompt`, `record_probe`) stays Python by design — same split as ports 1–2: pure logic to Rust, orchestration to Python. **Evidence (observed this session):** - **Parity 24/24 GREEN** (`parity_probe.py`): 17 schema-gate shapes + 3 non-dict top levels + 4 count_probes file shapes — byte-equal error lists pre-switch (vs the original vintage) and post-switch (round-trip through the shim). - **Consumer tests 23/23 GREEN** through the shim: `test_evidence_probe.py` (5) + `test_orchestration_evidence.py` + `test_runner_evidence.py` (18). - **Consumers verified:** `rune_operator.py:581/603` (`run_probe`, opt-in via MIMO_PROBE_ENABLED), `failure_recovery.py:40` (budget map), `research_stage_executor.py:312` (`run_probe` + `MAX_PROBES_PER_TASK` + `ProbeBudgetExceeded`). Import surface unchanged. **Parity traps proven this port (the silent-wrong class):** 1. **Non-string status double-error:** Python runs `status not in PROBE_STATUSES` on the RAW `.get` value — `{"status": 1}` fires BOTH the field error and the "must be one of … got 1" error. A naive Rust gate that skips the membership check for non-strings diverges silently (caught by the int-status battery case). 2. **Python repr rendering:** `got {status!r}` → single-quoted `'maybe'`; the sorted-set list renders `['CONTRADICTED', 'GROUNDED', 'STILL_UNKNOWN']` — comma-space, single quotes, sorted. 3. Python 3.14 dataclass law: an importlib-loaded module must be registered in `sys.modules` before `exec_module` or the `@dataclass` machinery raises `AttributeError: 'NoneType' has no attribute '__dict__'`. **Commits:** `9cb93e6` → `3450a1c`. Port order per contract §4: audit_ledger ✅ → verification_layer ✅ → evidence_probe ✅ → **behaviour_ledger (next)** → behavior → claim_audit.
conductor commented 2026-10-03 15:59:29 +01:00
Author
Owner
Copy link
Copy source

Port 4 (behaviour_ledger) — authority switch EXECUTED

Scope: src/rune/behaviour_ledger.py — the state machine (define / get / transitions / selective invalidation / queues / summary) served in-process by the Rust cdylib. File glue (load/save) + state constants stay Python; timestamps are pinned by the shim (now passed in) so the core is deterministic.

Evidence (observed this session):

  • Parity 13/13 GREEN (parity_behaviour.py): full-scenario battery — 3 behaviours defined, auto-verification, user-confirmed + other-confirmer real-world passes, selective invalidation with reasons override + owner-component hit, project/component filters, byte-equal error messages (duplicate define, unknown behaviour), save render (indent=1 sort_keys). Doc parity is FULL-document byte-equality.
  • Post-switch wall: constants intact (testing_mode.py:19 imports TEST_FAILED/USER_CONFIRMED at module load — verified), save/reload round-trip through the shim, selective invalidation live, iteration.py + testing_mode.py import chains green.
  • Consumers: iteration.py:23, testing_mode.py:19, rune_tui.py:3152 (lazy). Import surface unchanged.

Design notes:

  • Doc-in/doc-out FFI convention with {ok, err} envelope — the shim raises ValueError with the exact Python message on err. All three error classes byte-equal.
  • No clock in the Rust core — now is an argument everywhere. Parity becomes fully deterministic (the Python vintage's time.strftime was frozen in the harness to match).
  • Selective-invalidation semantics preserved exactly: only CONFIRMED_STATES records with dependency/owner overlap reopen; the default reason carries the Python sorted-list repr (['dep:one']).

Commits: 3450a1c → 8973caa. Port order per contract §4: audit_ledger ✅ → verification_layer ✅ → evidence_probe ✅ → behaviour_ledger ✅ → behavior (next) → claim_audit.

## Port 4 (behaviour_ledger) — authority switch EXECUTED **Scope:** `src/rune/behaviour_ledger.py` — the state machine (define / get / transitions / selective invalidation / queues / summary) served in-process by the Rust cdylib. File glue (load/save) + state constants stay Python; timestamps are pinned by the shim (`now` passed in) so the core is deterministic. **Evidence (observed this session):** - **Parity 13/13 GREEN** (`parity_behaviour.py`): full-scenario battery — 3 behaviours defined, auto-verification, user-confirmed + other-confirmer real-world passes, selective invalidation with reasons override + owner-component hit, project/component filters, byte-equal error messages (duplicate define, unknown behaviour), save render (indent=1 sort_keys). Doc parity is FULL-document byte-equality. - **Post-switch wall:** constants intact (`testing_mode.py:19` imports TEST_FAILED/USER_CONFIRMED at module load — verified), save/reload round-trip through the shim, selective invalidation live, `iteration.py` + `testing_mode.py` import chains green. - **Consumers:** `iteration.py:23`, `testing_mode.py:19`, `rune_tui.py:3152` (lazy). Import surface unchanged. **Design notes:** - Doc-in/doc-out FFI convention with `{ok, err}` envelope — the shim raises `ValueError` with the exact Python message on err. All three error classes byte-equal. - **No clock in the Rust core** — `now` is an argument everywhere. Parity becomes fully deterministic (the Python vintage's `time.strftime` was frozen in the harness to match). - Selective-invalidation semantics preserved exactly: only `CONFIRMED_STATES` records with dependency/owner overlap reopen; the default reason carries the Python sorted-list repr (`['dep:one']`). **Commits:** `3450a1c` → `8973caa`. Port order per contract §4: audit_ledger ✅ → verification_layer ✅ → evidence_probe ✅ → behaviour_ledger ✅ → **behavior (next)** → claim_audit.
conductor commented 2026-10-03 16:02:00 +01:00
Author
Owner
Copy link
Copy source

Port 5 (behavior) — authority switch EXECUTED

Scope: src/rune/behavior.py decision math (reinforce counting, confidence formulas, promotion/state thresholds) served in-process by the Rust cdylib (ep_habit_reinforce / ep_correction_reinforce). Storage stays Python — SQLite I/O and regex matching remain in the module: the crate is zero-dep by design, and hand-rolling SQLite or a regex engine would be hubris, not hardening. The deployed ~/.config/opencode/scripts/behavior.py vintage ALSO upgraded to the same shim shape (contract §2 vintage-upgrade) pointing at the LIVE db.

Evidence (observed this session):

  • Parity 16/16 GREEN (parity_behavior.py): 12 decision-math cases (outcome × state × boundaries: promotion-at-min-activations, confidence floors/ceilings, neutral counts activation, archive-threshold crossing) with SEMANTIC comparison (parse-then-compare — the 1.0 vs 1 delta is JSON rendering, not semantics) + 4 full-stack live-db scenarios running both sides' complete stacks (promotion observed, habit_activations logged identically).
  • Deployed shim live against the real db (~/.config/opencode/memory/behavior.db, 2 corrections / 1 habit — read-verified through the shim, data untouched; parity scenarios ran on temp dirs).
  • Consumers: rune_tui.py:83 (import behavior at startup) resolves the deployed shim; read paths (behavior_stats, match_context) verified against the live db.

Design decisions recorded:

  • The Python vintage's if new_count > 0: ... else: new_confidence = habit["confidence"] else-branch is DEAD CODE (activation_count+1 is always ≥ 1); the Rust core is faithful to the live path, and the dead branch stays dead.
  • No clock in the core — timestamps pinned by the shim (now passed in for the SQL UPDATE, not the math).
  • Deployed-vintage upgrade: same shim shape, live-db path — the src/rune authority vintage resolves its db via memory_env, the deployed copy via expanduser; both verified.

Commits: 8973caa → 2732677. Port order per contract §4: audit_ledger ✅ → verification_layer ✅ → evidence_probe ✅ → behaviour_ledger ✅ → behavior ✅ → claim_audit (next, last).

## Port 5 (behavior) — authority switch EXECUTED **Scope:** `src/rune/behavior.py` decision math (reinforce counting, confidence formulas, promotion/state thresholds) served in-process by the Rust cdylib (`ep_habit_reinforce` / `ep_correction_reinforce`). **Storage stays Python** — SQLite I/O and regex matching remain in the module: the crate is zero-dep by design, and hand-rolling SQLite or a regex engine would be hubris, not hardening. The deployed `~/.config/opencode/scripts/behavior.py` vintage ALSO upgraded to the same shim shape (contract §2 vintage-upgrade) pointing at the LIVE db. **Evidence (observed this session):** - **Parity 16/16 GREEN** (`parity_behavior.py`): 12 decision-math cases (outcome × state × boundaries: promotion-at-min-activations, confidence floors/ceilings, neutral counts activation, archive-threshold crossing) with SEMANTIC comparison (parse-then-compare — the `1.0` vs `1` delta is JSON rendering, not semantics) + 4 full-stack live-db scenarios running both sides' complete stacks (promotion observed, habit_activations logged identically). - **Deployed shim live against the real db** (`~/.config/opencode/memory/behavior.db`, 2 corrections / 1 habit — read-verified through the shim, data untouched; parity scenarios ran on temp dirs). - **Consumers:** `rune_tui.py:83` (`import behavior` at startup) resolves the deployed shim; read paths (`behavior_stats`, `match_context`) verified against the live db. **Design decisions recorded:** - The Python vintage's `if new_count > 0: ... else: new_confidence = habit["confidence"]` else-branch is DEAD CODE (activation_count+1 is always ≥ 1); the Rust core is faithful to the live path, and the dead branch stays dead. - No clock in the core — timestamps pinned by the shim (`now` passed in for the SQL UPDATE, not the math). - Deployed-vintage upgrade: same shim shape, live-db path — the src/rune authority vintage resolves its db via `memory_env`, the deployed copy via `expanduser`; both verified. **Commits:** `8973caa` → `2732677`. Port order per contract §4: audit_ledger ✅ → verification_layer ✅ → evidence_probe ✅ → behaviour_ledger ✅ → behavior ✅ → **claim_audit (next, last)**.
conductor commented 2026-10-03 16:19:22 +01:00
Author
Owner
Copy link
Copy source

Port 6 (claim_audit) — authority switch EXECUTED — ALL SIX PORTS DONE

Scope: src/rune/claim_audit.py — the decision skeleton (verify verdict tree, report aggregation, clean fold, evidence-term stoplist) served in-process by the Rust cdylib. Regex engine (scan/reframe/degrade/EVIDENCE_TAG), journal I/O, and ledger lookups stay Python (zero-dep crate law). BOTH copies shimmed (src/rune + deployed).

Evidence (observed this session):

  • Parity 17/17 GREEN (parity_claim.py): verify tree 12 cases (flags derived exactly as the original derives them, original-vs-FFI), report aggregate over a real-shaped journal incl. '?' fallback rows, clean fold 3 cases, stoplist filter.
  • Deployed shim live: the REAL claim_audit journal — 1127 rows (851 DEGRADE / 276 PASS) — aggregates through the Rust core; scan/audit/scan_text paths verified both sides.
  • Consumers: rune_tui.py:62 (import) + :2468 (report) + :3752 (scan_text), contracts.py dynamic imports, canary_native MODULES, hook_engine.py:270, shared/auto_detect.py:101 — import surface unchanged.

Two vintage-merge disclosures (this port is a MERGE, not just an upgrade):

  1. SYMMETRY GUARD preserved: the deployed scripts vintage carried research_lens.symmetry_check integration (2026-08-15) that src/rune's vintage LACKED — the shim serves it in BOTH no-claims and per-claim paths. Feature-preserving merge; the contract's "src/rune is authority" is honored for the shared surface, and the guard is load-bearing deployed behavior that would have been a regression to drop.
  2. DEAD CONSUMER REPAIRED: rune_tui.py:3752 calls claim_audit.scan_text(response[:500]) expecting {degraded: [...], claims: [...]} — NO implementation existed in EITHER vintage; the call died silently in try/except every TUI response. scan_text() is now implemented (audit → filtered dict), verified live returning the exact shape. This is a BUG FIX the port surfaced — the silent-try/except class, caught by reading the actual call site.

Commits: 2732677 → 969e39f. SLICE COMPLETE: all six ports executed (audit_ledger 114c394, verification_layer 9cb93e6, evidence_probe 3450a1c, behaviour_ledger 8973caa, behavior 2732677, claim_audit 969e39f). Slice-level close-out (§7): registry remap + regression gate + supersession ledger + final wall NEXT.

## Port 6 (claim_audit) — authority switch EXECUTED — ALL SIX PORTS DONE **Scope:** `src/rune/claim_audit.py` — the decision skeleton (verify verdict tree, report aggregation, clean fold, evidence-term stoplist) served in-process by the Rust cdylib. Regex engine (scan/reframe/degrade/EVIDENCE_TAG), journal I/O, and ledger lookups stay Python (zero-dep crate law). BOTH copies shimmed (src/rune + deployed). **Evidence (observed this session):** - **Parity 17/17 GREEN** (`parity_claim.py`): verify tree 12 cases (flags derived exactly as the original derives them, original-vs-FFI), report aggregate over a real-shaped journal incl. '?' fallback rows, clean fold 3 cases, stoplist filter. - **Deployed shim live:** the REAL claim_audit journal — 1127 rows (851 DEGRADE / 276 PASS) — aggregates through the Rust core; scan/audit/scan_text paths verified both sides. - **Consumers:** `rune_tui.py:62` (import) + `:2468` (report) + `:3752` (scan_text), `contracts.py` dynamic imports, `canary_native` MODULES, `hook_engine.py:270`, `shared/auto_detect.py:101` — import surface unchanged. **Two vintage-merge disclosures (this port is a MERGE, not just an upgrade):** 1. **SYMMETRY GUARD preserved:** the deployed scripts vintage carried `research_lens.symmetry_check` integration (2026-08-15) that src/rune's vintage LACKED — the shim serves it in BOTH no-claims and per-claim paths. Feature-preserving merge; the contract's "src/rune is authority" is honored for the shared surface, and the guard is load-bearing deployed behavior that would have been a regression to drop. 2. **DEAD CONSUMER REPAIRED:** `rune_tui.py:3752` calls `claim_audit.scan_text(response[:500])` expecting `{degraded: [...], claims: [...]}` — NO implementation existed in EITHER vintage; the call died silently in try/except every TUI response. `scan_text()` is now implemented (audit → filtered dict), verified live returning the exact shape. This is a BUG FIX the port surfaced — the silent-try/except class, caught by reading the actual call site. **Commits:** `2732677` → `969e39f`. **SLICE COMPLETE: all six ports executed** (audit_ledger `114c394`, verification_layer `9cb93e6`, evidence_probe `3450a1c`, behaviour_ledger `8973caa`, behavior `2732677`, claim_audit `969e39f`). Slice-level close-out (§7): registry remap + regression gate + supersession ledger + final wall NEXT.
conductor commented 2026-10-03 16:26:03 +01:00
Author
Owner
Copy link
Copy source

SLICE 2 EXECUTION COMPLETE — close-out + full-suite verification

Close-out commit 2e68b6b:

  • shared/audit_ledger.py DELETED — superseded (strangler step 11): zero importers verified by sweep, deployed shim + Rust core proven live before removal (language-governance ⑥).
  • Coverage registry remapped: audit_ledger removed from SHARED_D.
  • Supersession ledger: plans/2026-10-03-phaseG-evidence-provenance-supersession.md — port table (6/6 with commits, parity counts, consumer proofs), 12-step doctrine discharge, vintage-merge disclosures, live-data rules honored.
  • Regression gate: structural_coverage_audit exit 0 pre- AND post-deletion.

Full-suite verification (ci: register canary_audit_ledger commit 497d14a):

  • Full pytest suite: 2240 passed / 10 failed in 258s.
  • Failure triage (baseline discriminator — the same tests re-run at pre-slice commit 2ac2d0a in a clean worktree): 9 of 10 are PRE-EXISTING at the baseline (henxels contract module errors ×5, umbrella manifest, lint baseline, spine-rebuild neutrality — none touched by slice 2). 1 of 10 was MINE: test_every_canary_on_disk_is_registered_in_ci — the new canary wasn't in the CI loop (the henxels per-canary rule); FIXED in 497d14a (registered; the contract test doing exactly its job — gap #138's rule enforced live).
  • The canary itself: 34/34 GREEN (Python canary incl. port-2/3 sections); Rust canary 28/28.

Net slice-2 state: commits 2ac2d0a → 497d14a (7 commits); 6 ports executed; parity batteries 33+24+13+16+17 = 103 checks all green; canaries 34+28; regression gate green; branch pushed to both remotes; issue #2449 carries the evidence trail (comments 33817, 33819, 33820, 33821, 33822, 33823).

Remaining (per contract §7): none for EXECUTION — the slice is complete. T7.3 next steps are slice-3 selection (NOT part of this slice's scope).

## SLICE 2 EXECUTION COMPLETE — close-out + full-suite verification **Close-out commit `2e68b6b`:** - `shared/audit_ledger.py` DELETED — superseded (strangler step 11): zero importers verified by sweep, deployed shim + Rust core proven live before removal (language-governance ⑥). - Coverage registry remapped: `audit_ledger` removed from SHARED_D. - Supersession ledger: `plans/2026-10-03-phaseG-evidence-provenance-supersession.md` — port table (6/6 with commits, parity counts, consumer proofs), 12-step doctrine discharge, vintage-merge disclosures, live-data rules honored. - Regression gate: structural_coverage_audit exit 0 pre- AND post-deletion. **Full-suite verification (`ci: register canary_audit_ledger` commit `497d14a`):** - Full pytest suite: **2240 passed / 10 failed** in 258s. - Failure triage (baseline discriminator — the same tests re-run at pre-slice commit `2ac2d0a` in a clean worktree): **9 of 10 are PRE-EXISTING** at the baseline (henxels contract module errors ×5, umbrella manifest, lint baseline, spine-rebuild neutrality — none touched by slice 2). 1 of 10 was MINE: `test_every_canary_on_disk_is_registered_in_ci` — the new canary wasn't in the CI loop (the henxels per-canary rule); FIXED in `497d14a` (registered; the contract test doing exactly its job — gap #138's rule enforced live). - The canary itself: **34/34 GREEN** (Python canary incl. port-2/3 sections); Rust canary 28/28. **Net slice-2 state:** commits `2ac2d0a` → `497d14a` (7 commits); 6 ports executed; parity batteries 33+24+13+16+17 = 103 checks all green; canaries 34+28; regression gate green; branch pushed to both remotes; issue #2449 carries the evidence trail (comments 33817, 33819, 33820, 33821, 33822, 33823). **Remaining (per contract §7):** none for EXECUTION — the slice is complete. T7.3 next steps are slice-3 selection (NOT part of this slice's scope).
conductor commented 2026-10-04 02:25:00 +01:00
Author
Owner
Copy link
Copy source

Slice-2 close-out — evidence (correcting 33826)

The completion comment posted at 16:26 (33826) was one minute premature: CI on 497d14a came back red (canary job failed; the new lint/type/lock gates also red). This comment supersedes it with observed evidence. Root causes, honestly split:

Slice-2's own debt (mine to fix):

  • canary completeness (frozen contract §5/§10): the canary pair carried ports 1-2 only. Now both canaries hold per-port sections for ALL SIX ports — canary_audit_ledger.py 74/74, canary_evidence_provenance_rust.py 66/66.
  • CI-safety: the audit-ledger canary pinned live-ledger truths that don't exist in CI's sandboxed HOME. Two repair layers: a DUAL-MODE law (LIVE on desktop / ISOLATED in CI — the discriminator is RUNE_HOME=/tmp/rune-ci-*, NOT file existence: run 18141's shim writes left a zombie 1-entry ledger in the runner's persistent HOME that flipped a naive isfile() gate on run 18161, now fixed and documented), and a ghost-divergence repair (the Rust binary honors RUNE_MEMORY_DIR first while the shim resolved $HOME — both surfaces now always resolve the SAME ledger; §5 strips RUNE_* from binary envs).
  • diff-cover 70% floor: the strangler moved shims' verification into canaries (outside pytest), so changed lines measured 46%. New tests/test_evidence_provenance_shims.py drives the shim delegation surfaces directly (6 tests) → 87% measured locally.
  • artifact coupling: canary resolved the binary from an absolute desktop path; now repo-relative with EP_RUST_LIB override (file-form, matching ep_loader's convention).

Pre-existing latent debt, first exposed by the new CI profile (fixed in passing):

  • umbrella manifest: ffi_query_memory_smoke.py unmapped since 43e7f5f (umbrella gate green at the true merge-base 9425ccc with UMBRELLA_SKIP_PROPAGATION=1).
  • dependency lock drift: markupsafe 3.0.3→3.0.4, websockets 17.1→17.2 (regenerated, checker green; lock law held — regenerated then committed, never auto-mutated).
  • lint baseline: claim_audit's 4 deliberate-silence handlers repaired with exc_info logging + reasoned noqa (BLE001 347→345, honest shrink); canary battery handlers carry the canary-law reason.
  • type ratchet warnings 454→457 (pyright counted the new shims): 3 runtime-shape isinstance checks carry type: ignore (legitimate runtime canary checks); the ep_loader imports carry the import-not-found annotation.
  • ffi_query_memory_smoke coverage: the smoke was stranded (imported the old deleted Python vintage); repaired to drive the shared/ shim (41/41 green) + tests/test_ffi_query_memory_smoke.py added. Also plan filenames renamed kebab-case per henxels rule.
  • deterministic type instrument (repair, not reading-management): ratchet now pins pyright --pythonpath to the invoking venv. Ambient-interpreter resolution was the residual non-determinism (desktop 454 vs runner 460 on the identical tree; runner's py3.11 lacks rich/Pillow, and import-level ignores flip to unnecessary-ignores across Python versions). Pillow is now declared (real lazy dep). Measured under the pinned instrument: 115 errors / 439 warnings (CI-equiv locked 3.11 venv) and 115/436 (local 3.14) — both below the 117/454 baseline.
  • Rust toolchain law: system rustc (pacman rust 1:1.99.0-1) is silently-broken — librustc_driver.so references narrow-char std::string::_M_mutate under the LLVM_23.1 version node that /usr/lib/libLLVM.so.23.1 doesn't define there (only wide-char _M_mutate carries it). cargo --version lies; every compile dies. CI now points PATH at the healthy rustup toolchain (which ships its own libLLVM; /home/nick carries the runner-user x-ACL) and builds to a runner-owned CARGO_TARGET_DIR (the repo's target/ is nick-owned).

Observation on the runner's persistent HOME (maintenance-class, not blocking): run 18141's canary writes left a zombie 1-entry promotion_ledger.jsonl in /var/lib/gitea-desktop-runner/.config/opencode/memory/. CI's own job env now overrides RUNE_HOME/RUNE_MEMORY_DIR, so it no longer affects runs, but the zombie file merits cleanup.

Final wall (all observed on this machine):

  • 5 parity harnesses: parity_verification 33/33, parity_probe 24/24, parity_behaviour 13/13, parity_behavior 16/16, parity_claim 17/17
  • canary pair: 74/74 + 66/66 in LIVE mode desktop; 74/74 + 66/66 in ISOLATED mode under simulated runner topology (zombie ledger present, RUNE_HOME set) — after run 18161's flake was traced, mode detection was fixed, and 10 consecutive CI-mode runs plus 4 LIVE/ISOLATED alternating runs passed clean
  • pytest suite: 2257 passed, 0 failed (includes the new shim battery + smoke coverage test)
  • consumer chains: iteration, testing_mode, rune_operator, contracts import through the shims; deployed curator + 4 shims serve the live surface; live chain intact (entries 35)
  • gates: lint ratchet 345/75 PASS; type ratchet 115/439 PASS (CI-equiv); dep-lock in sync; henxels exit 0; umbrella exit 0 (push-mode)

CI on 02a9dc4: run 18172 — all 9 jobs green (lint, unit-tests, canary, best-practices, documentation, type-check, gitea-offline, required-checks, notify-event-bridge).

Slice-2 close-out is now honestly complete: six ports parity-proven, both remotes pushed, CI green with the full canary pair carrying all six per-port sections.

## Slice-2 close-out — evidence (correcting 33826) The completion comment posted at 16:26 (`33826`) was one minute premature: CI on `497d14a` came back red (canary job failed; the new lint/type/lock gates also red). This comment supersedes it with observed evidence. Root causes, honestly split: **Slice-2's own debt (mine to fix):** - canary completeness (frozen contract §5/§10): the canary pair carried ports 1-2 only. Now both canaries hold per-port sections for ALL SIX ports — `canary_audit_ledger.py` 74/74, `canary_evidence_provenance_rust.py` 66/66. - CI-safety: the audit-ledger canary pinned live-ledger truths that don't exist in CI's sandboxed HOME. Two repair layers: a DUAL-MODE law (LIVE on desktop / ISOLATED in CI — the discriminator is `RUNE_HOME=/tmp/rune-ci-*`, NOT file existence: run 18141's shim writes left a zombie 1-entry ledger in the runner's persistent HOME that flipped a naive isfile() gate on run 18161, now fixed and documented), and a ghost-divergence repair (the Rust binary honors RUNE_MEMORY_DIR first while the shim resolved $HOME — both surfaces now always resolve the SAME ledger; §5 strips RUNE_* from binary envs). - diff-cover 70% floor: the strangler moved shims' verification into canaries (outside pytest), so changed lines measured 46%. New tests/test_evidence_provenance_shims.py drives the shim delegation surfaces directly (6 tests) → 87% measured locally. - artifact coupling: canary resolved the binary from an absolute desktop path; now repo-relative with EP_RUST_LIB override (file-form, matching ep_loader's convention). **Pre-existing latent debt, first exposed by the new CI profile (fixed in passing):** - umbrella manifest: ffi_query_memory_smoke.py unmapped since 43e7f5f (umbrella gate green at the true merge-base 9425ccc with UMBRELLA_SKIP_PROPAGATION=1). - dependency lock drift: markupsafe 3.0.3→3.0.4, websockets 17.1→17.2 (regenerated, checker green; lock law held — regenerated then committed, never auto-mutated). - lint baseline: claim_audit's 4 deliberate-silence handlers repaired with exc_info logging + reasoned noqa (BLE001 347→345, honest shrink); canary battery handlers carry the canary-law reason. - type ratchet warnings 454→457 (pyright counted the new shims): 3 runtime-shape isinstance checks carry type: ignore (legitimate runtime canary checks); the ep_loader imports carry the import-not-found annotation. - ffi_query_memory_smoke coverage: the smoke was stranded (imported the old deleted Python vintage); repaired to drive the shared/ shim (41/41 green) + tests/test_ffi_query_memory_smoke.py added. Also plan filenames renamed kebab-case per henxels rule. - deterministic type instrument (repair, not reading-management): ratchet now pins pyright --pythonpath to the invoking venv. Ambient-interpreter resolution was the residual non-determinism (desktop 454 vs runner 460 on the identical tree; runner's py3.11 lacks rich/Pillow, and import-level ignores flip to unnecessary-ignores across Python versions). Pillow is now declared (real lazy dep). Measured under the pinned instrument: 115 errors / 439 warnings (CI-equiv locked 3.11 venv) and 115/436 (local 3.14) — both below the 117/454 baseline. - Rust toolchain law: system rustc (pacman rust 1:1.99.0-1) is silently-broken — librustc_driver.so references narrow-char std::string::_M_mutate under the LLVM_23.1 version node that /usr/lib/libLLVM.so.23.1 doesn't define there (only wide-char _M_mutate carries it). cargo --version lies; every compile dies. CI now points PATH at the healthy rustup toolchain (which ships its own libLLVM; /home/nick carries the runner-user x-ACL) and builds to a runner-owned CARGO_TARGET_DIR (the repo's target/ is nick-owned). **Observation on the runner's persistent HOME** (maintenance-class, not blocking): run 18141's canary writes left a zombie 1-entry promotion_ledger.jsonl in /var/lib/gitea-desktop-runner/.config/opencode/memory/. CI's own job env now overrides RUNE_HOME/RUNE_MEMORY_DIR, so it no longer affects runs, but the zombie file merits cleanup. Final wall (all observed on this machine): - 5 parity harnesses: parity_verification 33/33, parity_probe 24/24, parity_behaviour 13/13, parity_behavior 16/16, parity_claim 17/17 - canary pair: 74/74 + 66/66 in LIVE mode desktop; 74/74 + 66/66 in ISOLATED mode under simulated runner topology (zombie ledger present, RUNE_HOME set) — after run 18161's flake was traced, mode detection was fixed, and 10 consecutive CI-mode runs plus 4 LIVE/ISOLATED alternating runs passed clean - pytest suite: 2257 passed, 0 failed (includes the new shim battery + smoke coverage test) - consumer chains: iteration, testing_mode, rune_operator, contracts import through the shims; deployed curator + 4 shims serve the live surface; live chain intact (entries 35) - gates: lint ratchet 345/75 PASS; type ratchet 115/439 PASS (CI-equiv); dep-lock in sync; henxels exit 0; umbrella exit 0 (push-mode) CI on 02a9dc4: run 18172 — **all 9 jobs green** (lint, unit-tests, canary, best-practices, documentation, type-check, gitea-offline, required-checks, notify-event-bridge). Slice-2 close-out is now honestly complete: six ports parity-proven, both remotes pushed, CI green with the full canary pair carrying all six per-port sections.
conductor commented 2026-10-08 15:18:53 +01:00
Author
Owner
Copy link
Copy source

#2449 progress note — grounding census + batch A landed (the issue stays OPEN for the remaining batches).

Grounding census (contract BUILDER-2026-10-08T14:02:24Z-588ca6, closed): docs/phaseG_slice2_census.md — the measured per-module state: claim_audit 17/17 SWITCHED; behavior/behaviour_ledger/evidence_probe/verification_layer loader-integrated; audit_ledger the one gap — with the drift-key pin and vintage-reconciliation constraint as frozen-contract input. Batteries green: 14/14 + 6/6 shims.

Batch A — the audit_ledger switch (contract BUILDER-2026-10-08T14:06:45Z-f1eada, closed):

  • The defect: SECRET_KEY defaulted to rune-governed-adaptation-2024 while the LIVE 62-entry chain verifies under the DEPLOYED vintage rune-drift-integrity-2024 — proved live (deployed: 62/valid; old default: 0 entries). A src/rune append would have corrupted the chain under the authority key. The key ADOPTS the authority (RECONCILE-DON'T-CLOBBER: one chain, one key).
  • verify_chain + _compute_hash delegate to the ep seam (Rust authority); no silent fallback — a missing cdylib fails loudly. Rust verify on the live ledger: 62/valid.
  • tests/test_audit_ledger_switch.py 5/5: key adoption, Rust-vs-Python HMAC parity, append→verify round-trip, tamper detection through Rust, live ledger under the authority key. Both import orders green; family 12/12; lint ratchet PASS.

Remaining per the census: vintage reconciliation, supersession ledger, ep-shadow evidence.

#2449 progress note — grounding census + batch A landed (the issue stays OPEN for the remaining batches). **Grounding census** (contract BUILDER-2026-10-08T14:02:24Z-588ca6, closed): docs/phaseG_slice2_census.md — the measured per-module state: claim_audit 17/17 SWITCHED; behavior/behaviour_ledger/evidence_probe/verification_layer loader-integrated; audit_ledger the one gap — with the drift-key pin and vintage-reconciliation constraint as frozen-contract input. Batteries green: 14/14 + 6/6 shims. **Batch A — the audit_ledger switch** (contract BUILDER-2026-10-08T14:06:45Z-f1eada, closed): - The defect: SECRET_KEY defaulted to `rune-governed-adaptation-2024` while the LIVE 62-entry chain verifies under the DEPLOYED vintage `rune-drift-integrity-2024` — proved live (deployed: 62/valid; old default: 0 entries). A src/rune append would have corrupted the chain under the authority key. The key ADOPTS the authority (RECONCILE-DON'T-CLOBBER: one chain, one key). - verify_chain + _compute_hash delegate to the ep seam (Rust authority); no silent fallback — a missing cdylib fails loudly. Rust verify on the live ledger: 62/valid. - tests/test_audit_ledger_switch.py 5/5: key adoption, Rust-vs-Python HMAC parity, append→verify round-trip, tamper detection through Rust, live ledger under the authority key. Both import orders green; family 12/12; lint ratchet PASS. Remaining per the census: vintage reconciliation, supersession ledger, ep-shadow evidence.
gitea-actions Bot added the ready label 2026-10-08 20:10:06 +01:00
gitea-actions Bot commented 2026-10-08 20:10:06 +01:00
Copy link
Copy source

🎓 NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3) — dependencies satisfied (scheduled sweep); Rune may start.

🎓 **NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3)** — dependencies satisfied (scheduled sweep); Rune may start.
gitea-actions Bot commented 2026-10-08 20:35:05 +01:00
Copy link
Copy source

🎓 NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3) — dependencies satisfied (scheduled sweep); Rune may start.

🎓 **NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3)** — dependencies satisfied (scheduled sweep); Rune may start.
gitea-actions Bot commented 2026-10-08 20:36:52 +01:00
Copy link
Copy source

🎓 NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3) — dependencies satisfied (closed: #2478); Rune may start.

🎓 **NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3)** — dependencies satisfied (closed: #2478); Rune may start.
gitea-actions Bot commented 2026-10-08 20:36:55 +01:00
Copy link
Copy source

🎓 NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3) — dependencies satisfied (closed: #258); Rune may start.

🎓 **NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3)** — dependencies satisfied (closed: #258); Rune may start.
gitea-actions Bot commented 2026-10-08 20:36:57 +01:00
Copy link
Copy source

🎓 NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3) — dependencies satisfied (closed: #259); Rune may start.

🎓 **NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3)** — dependencies satisfied (closed: #259); Rune may start.
gitea-actions Bot commented 2026-10-08 20:36:59 +01:00
Copy link
Copy source

🎓 NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3) — dependencies satisfied (closed: #16); Rune may start.

🎓 **NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3)** — dependencies satisfied (closed: #16); Rune may start.
gitea-actions Bot commented 2026-10-08 20:37:00 +01:00
Copy link
Copy source

🎓 NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3) — dependencies satisfied (closed: #175); Rune may start.

🎓 **NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3)** — dependencies satisfied (closed: #175); Rune may start.
gitea-actions Bot commented 2026-10-08 20:37:02 +01:00
Copy link
Copy source

🎓 NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3) — dependencies satisfied (closed: #181); Rune may start.

🎓 **NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3)** — dependencies satisfied (closed: #181); Rune may start.
gitea-actions Bot commented 2026-10-08 20:42:52 +01:00
Copy link
Copy source

🎓 NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3) — dependencies satisfied (closed: #356); Rune may start.

🎓 **NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3)** — dependencies satisfied (closed: #356); Rune may start.
gitea-actions Bot commented 2026-10-08 20:42:54 +01:00
Copy link
Copy source

🎓 NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3) — dependencies satisfied (closed: #398); Rune may start.

🎓 **NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3)** — dependencies satisfied (closed: #398); Rune may start.
gitea-actions Bot commented 2026-10-08 20:42:56 +01:00
Copy link
Copy source

🎓 NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3) — dependencies satisfied (closed: #403); Rune may start.

🎓 **NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3)** — dependencies satisfied (closed: #403); Rune may start.
gitea-actions Bot commented 2026-10-08 20:42:58 +01:00
Copy link
Copy source

🎓 NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3) — dependencies satisfied (closed: #404); Rune may start.

🎓 **NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3)** — dependencies satisfied (closed: #404); Rune may start.
gitea-actions Bot commented 2026-10-08 20:43:00 +01:00
Copy link
Copy source

🎓 NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3) — dependencies satisfied (closed: #405); Rune may start.

🎓 **NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3)** — dependencies satisfied (closed: #405); Rune may start.
gitea-actions Bot commented 2026-10-08 20:43:02 +01:00
Copy link
Copy source

🎓 NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3) — dependencies satisfied (closed: #406); Rune may start.

🎓 **NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3)** — dependencies satisfied (closed: #406); Rune may start.
gitea-actions Bot commented 2026-10-08 20:43:04 +01:00
Copy link
Copy source

🎓 NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3) — dependencies satisfied (closed: #413); Rune may start.

🎓 **NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3)** — dependencies satisfied (closed: #413); Rune may start.
gitea-actions Bot commented 2026-10-08 20:43:06 +01:00
Copy link
Copy source

🎓 NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3) — dependencies satisfied (closed: #414); Rune may start.

🎓 **NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3)** — dependencies satisfied (closed: #414); Rune may start.
gitea-actions Bot commented 2026-10-08 20:43:08 +01:00
Copy link
Copy source

🎓 NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3) — dependencies satisfied (closed: #415); Rune may start.

🎓 **NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3)** — dependencies satisfied (closed: #415); Rune may start.
gitea-actions Bot commented 2026-10-08 20:43:10 +01:00
Copy link
Copy source

🎓 NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3) — dependencies satisfied (closed: #411); Rune may start.

🎓 **NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3)** — dependencies satisfied (closed: #411); Rune may start.
gitea-actions Bot commented 2026-10-08 20:45:02 +01:00
Copy link
Copy source

🎓 NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3) — dependencies satisfied (scheduled sweep); Rune may start.

🎓 **NEXT: #2449 — Phase G strangler slice 2: evidence-provenance Rust candidate (T7.3)** — dependencies satisfied (scheduled sweep); Rune may start.
gitea-actions Bot added the P2 — mediumgovernancegrowth labels 2026-10-08 21:17:46 +01:00
Sign in to join this conversation.
No Branch/Tag Specified
Labels
Clear labels
P0 — critical
Spine/core correctness or blocks other work — do first
 P1 — high
Material user value or real defect (money, bugs, next-stage dep)
 P2 — medium
Quality/consolidation improvement
 P3 — low
Measurement / benchmark / deferred policy
 blocked
Dependencies open
 bug
Confirmed defect or regression
 custodian
epic
Rune v1 course-correction issue graph
 feature
New capability or improvement
 governance
growth
identity
in-progress
Currently being worked on by Rune
 incomplete
Incomplete work — draft PR with ledger, awaiting resume
 infrastructure
life
memory
memory-atlas
Memory Research Atlas child domain
 needs-testing
Needs manual testing with dumbed-down steps before close
 perception
ready
Order satisfied - Rune may start
 release-drift
research
Research jobs — spikes, watchlists, evidence work
 rune
rune
 scaffolding
scaffolding
 spine,correction
spine,correction
 spine,deferred
spine,deferred
 spine,resolved
spine,resolved
 strangler,burn,endgame,deferred
strangler,burn,endgame,deferred
 strangler,burn,endgame,inventory,epic
strangler,burn,endgame,inventory,epic
 tui
tui
 tui,epic
tui,epic
 tui,openspec
tui,openspec
 v1-course-correction
Rune v1 course-correction issue graph
No labels P2 — medium governance growth ready
Milestone
No items
No Milestone
Projects
Clear projects
No projects
Assignees
Clear assignees
conductor
No Assignees
2 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: conductor/rune#2449