[mchwang]

Summary

Documentation only. The status records still described the pre-merge automation (#22) as unfinished and lane F as the next step. F3–F6 have merged, T9 is complete (#178), and #22 is closed.

• README.md: the status line lists the one-click merge job. With a runner, the Merge PR button pushes the rebased head, waits for required checks and hands the exact pair to the guarded merge.
• docs/architecture.md:
  • The status line moves to eafe2be (2026-10-20). "Partly built for pre-merge automation" becomes "In use".
  • The lane F row names F6a–F6c (PRs F6a: push the rewritten head behind a durable marker #153, F6b: one-click merge job with a required-check wait #161, F6c: run every T9 suite in CI and add the live probe workflow #177), T9 (PR Record that T9 is complete #178), the path-identity probe (Probe checkout path identity at runner startup #162, PR Probe task storage path identity at runner startup (#162) #180), the open follow-ups (Keep the rebase conflict resolver's trust-read timeout over a later stop #174, App-pinned required checks always read as missing #176, Host checkouts on a case-insensitive host merge paths (staging clone, rebase worktree) #179, Sweep leftover path identity probe containers in startup recovery #181), and notes that T6 and T12 still have Verify gaps.
  • The button is called Merge PR (the design's Approve & merge). Flow 3 now describes the runner's merge job.
• docs/designs/codeboost-plan-indexed-review.md:
  • Milestone 5: the runner has merged; review rounds and runner buttons are not built.
  • Milestone 4 is mostly done, not done. The same open-item list appears in all three places: the T6 and T12 Verify gaps, Keep the rebase conflict resolver's trust-read timeout over a later stop #174, App-pinned required checks always read as missing #176, and the merge job's full step list and Cancel button.
  • Build step 4 checklist: the rebase and cmd: items are ticked with their PRs; "validate the complete pre-merge sequence" stays open.
  • Implementation Tasks: T3, T7 and T11 are ticked with test evidence (a clause-by-clause audit of each Verify line). T6 and T12 stay unticked, each with a status line. T6 has no test that cmd: checks run again on a head-only collaborator push, or that rule sources are read again when SHAs match. T12 uses hand-written output rather than recorded gh output, and App-pinned required checks always read as missing #176 breaks app-pinned checks.

Tracking issues closed earlier today, each with an evidence comment: #28 (lane D), #29 (lane E), #66 (PRs #85, #90), #88 (PR #134).

Validation

• git diff --check passes on every commit. No test or script reads these docs.
• Every cited PR and issue number was checked against GitHub. All 19 cited test titles exist word for word, and their bodies were read.

Review record

Base eafe2be throughout. Each independent pass was a fresh general-purpose subagent reviewing the full diff, with no author context.

Round	Head	Findings	What changed
Independent 1	1ef15ed	5: "every T9 suite on every PR" overstates (live probes are nightly); lane F listed done though review rounds are not built; sequencing paragraph still pending; README button name; chained "and"	All fixed in 00816bb
Independent 2	00816bb	4 + nits: T3/T6/T7/T11/T12 unticked while step 4 said Done; arch "no buttons" vs Merge PR; Approve & merge naming; missing step list/Cancel; F6a PR number; tense	Clause-by-clause test audit: T3/T7/T11 ticked with evidence, T6/T12 kept open with gaps, step 4 changed to "mostly done" (06dbcdf)
Independent 3	06dbcdf	3: sequencing paragraph said step 4 complete; stale "current work" line; T7 base-race evidence	Fixed in aaa6264
Independent 4	aaa6264	3: replacing the refusal is runner-only; T6 evidence lacked the no-merge test; T7 head-only test	Fixed in ac8afcb
Independent 5	ac8afcb	3: step 4 open-item lists differed; Flow 3 described only the direct merge; browser also sends an action ID	Fixed in c2a8d01
Independent 6	c2a8d01	3 wording: T11 fake-resolver caveat, T12 output source, T7 refusal caveat	Fixed in 4ce6605; for T7, simulated base refusals cited instead of the suggested wording, because they exist (merge-job.test.ts, pre-merge.test.ts)
Independent 7	4ce6605	1: T11 caveat missed ledger-seeded tests	Fixed in a404a0e
Independent 8	a404a0e	No findings	Clean pass; authorizes the Copilot request on a404a0e

Nothing was declined.

Review-lesson audit

• Overstated or stale status claims (rounds 1–7): covered by the existing rule "Treat every behavioural claim the change makes … as something to verify" (CLAUDE.md/AGENTS.md, Review readiness).
• Ticking T items without Verify-clause evidence (round 2): covered by the design's own rule "An increment must not mark a broader requirement complete while any of its acceptance criteria remain deferred".
• No new AGENTS.md rule is needed.

Deferred follow-ups

• The T6 test gaps and the T12 recorded-output gap have no issue yet. App-pinned required checks always read as missing #176 tracks the app-pinned check bug.

🤖 Generated with Claude Code

[Copilot]

🟢 Approval recommended

The documentation-only changes reflect completed work and explicitly retain outstanding limitations, with no unresolved findings.

No open issues 🎉

What changed in this PR

Updates documentation to reflect completed pre-merge automation while preserving known limitations and verification gaps.

Changes:

• Documents the runner’s one-click Merge PR workflow.
• Refreshes lane F, milestone, and closed-issue status.
• Records T3, T7, and T11 evidence while keeping T6 and T12 open.

File	Description
README.md	Updates the feature status and merge workflow.
docs/​designs/​codeboost-plan-indexed-review.md	Refreshes milestones, task evidence, and remaining gaps.
docs/​architecture.md	Updates runner status, merge flow, and follow-ups.

_{Review effort: Balanced}