Skip to content
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
33 changes: 33 additions & 0 deletions docs/archive/reviews/2026-07-06-p04-p12-target-state-review.md
Original file line number Diff line number Diff line change
Expand Up @@ -355,3 +355,36 @@ all four were not delivered.
coordinator's reported failing number two branches below it. Both are consistent with a gate
that has effectively zero headroom; neither number should be treated as the authoritative
figure without a fresh CI run.

## 7. Closure resolution log (current `main` after PR #79)

> **Added 2026-07-06 as a separated closeout log.** The audit above stays pinned to
> `a239986`. This section records the later remediation state verified on current `main` at
> `98993ff` (`test: capture evrun full smoke evidence (#79)`), using only post-audit repo
> evidence now present on `main`.

| ID | Status | Current-main closure evidence |
| --- | ---------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| F1 | fixed | PR #72 / `0dfb9d0` (`test: restore branch-coverage headroom above the gate floor`) added the missing branch-coverage headroom back into the test corpus, and current [`verification.md`](../../delivery/target-state-implementation/verification.md) now codifies both the 90.5% headroom rule and the post-merge watch that this finding said were missing. |
| F2 | fixed | PR #74 / `81d5664` (`fix: support live parked decisions`) rewired the live parked decision path in `sdk.ts`, `harness.ts`, `projection.ts`, and `resume.ts`; current tests include `P09-F2: decide routes a live parked run instead of refusing` in `packages/jig-sdk/tests/sdk.unit.test.ts`. |
| F3 | partially closed | PR #78 / `dcffa56` (`docs: record p05 real forge smoke rerun`) added a current-checkout real Forge/work-source rerun under [`docs/design/evidence/2026-07-06-p05-real-forge-smoke-rerun.md`](../../design/evidence/2026-07-06-p05-real-forge-smoke-rerun.md) and updated front-door status docs to qualify the shipped claim. The same evidence and current `README.md`/`AGENTS.md`/`docs/README.md` remain explicit that held-merge plus commit-status/comment block-surfacing real-effect proof is still open, so `MERGE-5` is not fully closed. |
| F4 | fixed | PR #79 / `98993ff` updated P11 from an internally contradictory blocked-capture posture to a reconciled combined-smoke posture in [`phases/11-evrun-full-evidence.md`](../../delivery/target-state-implementation/phases/11-evrun-full-evidence.md), with per-AC current dispositions that distinguish what is now met from what remains stronger-probe debt. The finding's contradiction is closed without overclaiming no-phone-home or idempotency. |
| F5 | fixed | PR #77 / `a78f67c` (`docs: harden target-state delivery tracking`) corrected the stale track header, phase-table rows, and P11/P12 status lines on current `main`, and tightened the delivery-foundation/status-flip checks so the daisy-chain failure mode is no longer the documented workflow. |
| F6 | fixed | PR #77 / `a78f67c` updated [`docs/design/contracts/driving.md`](../../design/contracts/driving.md) to the shipped 10-verb operator-control vocabulary (`start`, `preview`, `watch`, `inspect`, `ask-why`, `notice-ack`, `notice-snooze`, `decide`, `stop`, `export`), matching ADR 0033 and the current `JigOperatorControlPort`. |
| F7 | fixed | PR #77 / `a78f67c` removed the stale "resume placement is unsettled" posture from the track README and now states the settled split directly: resume lives on `JigRecoverySurface`, not on the operator-control port or MCP surface. |
| F8 | fixed | PR #75 / `1e23b5f` (`fix: exercise negative egress probe`) replaced the hardcoded `negativeEgressProbePassed: false` behavior with an exercised probe result path in `packages/jig-sdk/src/providers/real/confinement.ts`; current Phase 04 post-merge notes record that remediation explicitly. |
| F9 | fixed | PR #75 / `1e23b5f` also closed the latent substrate-validation hole: current `createRealExecutionHost()` now refuses probe substrate requests when no approved manifest is supplied and validates every declared request before exercising the probe. |
| F10 | fixed | PR #76 / `a66747d` (`feat: surface export attribution in observations`) added export attribution to inspect/ask-why and covered it with current CLI and SDK tests, including `P10-F10: custom-dir export remains attributable from inspect and run-level ask-why`. |
| F11 | fixed | PR #76 / `a66747d` added the missing second-process watch integration test (`P08-F11: watch observes a live run from a second process and then the finished record`) and updated the phase doc to state the shipped watch vocabulary honestly: LIVE posture is projected into snapshot signals rather than a literal `stalled` status string. |
| F12 | fixed | PR #77 / `a78f67c` added the missing process teeth to current delivery rules: explicit review-arrival gate, post-merge main-branch check watch, 90.5% coverage headroom, and a rewritten same-PR `merged (#N)` status-flip rule in [`verification.md`](../../delivery/target-state-implementation/verification.md). The historical incidents remain part of the audit record, but the process gap identified by F12 is now addressed in the governing docs. |

### Remaining open debt after the remediation wave

- F3 is only partially closed. Current docs and the P05 rerun record still say the success-path
real Forge `open-pr` smoke is proven, while held-merge and commit-status/comment block-surfacing
legs remain unit-proven or otherwise unevidenced on a real-effect path.
- P11 evidence is stronger than it was at `a239986`, but it is still not a claim that
no-phone-home or repeated-effect idempotency is proven. Current P11 docs, evidence index, root
`README.md`, repo `AGENTS.md`, and track `verification.md` all continue to treat those as open
evidence or owner-deferral work, so this closeout does not advance P13/P14 beyond what current
`main` already states.