[pascalandy]

Outside the plan: Pascal assigned #412 and #413, from the UATs of #410 and #411. This pull request follows #411, merged at 530f2de.

The UATs found gaps in just finish, the exit gate of the loop of the nicely-extension skill. An indented Fill: line passed. A missing form stopped finish on open qa.md: no such file or directory, with no fix and no other problem. A warn finding passed beside an advise when it sat in a table after a valid ### finding, in fenced code, or in prose. Finish read Reviewer: inside fenced code. And no delivery step ran finish, so a "full stop" request could end the rounds early.

Scope

• Forms: finish reads a Fill: line after its indent. It reports a missing form with a fix, then goes on to the other checks (formProblems in sdk/extensionrepo/finish.go)
• Findings: under ## Findings, each line that names a severity fails finish, except the bullet - **Severity**: of a finding. That covers the word severity, a severity in code, and a severity cell of a table, fenced code included. A heading of level 1 or 2 closes the findings first (readinessFindings in readiness.go)
• Header: finish reads Commit:, Reviewer:, and Verdict: only in the first three lines of readiness.md, as the template skill says
• Signoff: an extension repository runs just finish, then just check, in the scratch worktree of just signoff. The release needs that signoff, so it waits too. Core keeps just check alone (run in program.go, signoff.go)
• Docs: D016, extension-repository.md, the template AGENTS.md, and the template skill. The rule index now names signoff and the finish gate for "Stop at Done when" and "Each round ends with an independent review"
• Three moves keep files under the limit of just limits, with no change of behavior: the helpers of the finish test, the signoff test, and the rule index, now AGENTS/rule-index.md of the template

Proved by

Each new case of scripts/template-finish-test.sh and scripts/template-signoff-test.sh failed first on the code before its fix, then passed:

• Indented Fill:: want present "vision.md:19: this line of the form is not filled"
• Missing form: finish: open spec.md: no such file or directory
• Severity in a table, fenced code, a plain bullet, and prose: just finish exited 0, want 1
• Reviewer: in fenced code: just finish exited 0, want 1
• Signoff of a repository whose finish fails: Signed off on c7d4ffd0...
• Review findings S1 (### Severity: warn) and S2 (## Severity assessment): want present "readiness.md:33: ..." and just finish exited 1, want 0

The cases of an unknown commit, a misspelled verdict, and a commit after the review cover refusals that already worked. A mutation of each refusal in readiness.go made its case fail.

• 1. Spec: n/a: no ncly command, flag, JSON key, or error code changes. extension-repository.md lists the new checks of just finish and just signoff
• 2. Scenario: each case above fails on the code before its fix
• 3. Code: those cases pass
• 4. Text: n/a: just finish and just signoff print repository-task diagnostics, and no catalog entry changes
• 5. Checks: just check passes at eb6a57c8 on Linux (om1), and on macOS (mini) through check_remote.py with the default TMPDIR: "ok":true, "exit_code":0
• 6. Verify: GPT-6 Luna on xhigh drove the template feature at eb6a57c85c9f, not dirty. Every step passed. Then it ran probes on a copy of the rendered repository:
  • A row | Output keeps GPS | `warn` | after a valid advise finding: exit 1, with readiness.md:11: this line names a severity outside the line - **Severity**: of a finding
  • The same review with ## Severity assessment in place of the row: exit 0
  • ### Severity: warn beside an advise: exit 1 at its line, and no problem for ## Severity assessment
  • Reviewer: only in fenced code: exit 1, with readiness.md: it names no reviewer
  • A run at f73f6500 proved fix(template): just finish : détecter un Fill: indenté, guider quand un formulaire manque, tester les refus #412: vision.md:19: this line of the form is not filled for an indented line, and qa.md: the form is missing. Fix: restore it from the Git history of the repository, then fill it beside the other problems. finish.go has not changed since
• 7. Docs: no backlog item, since both issues are outside the plan. The template skill teaches where a severity goes, and the template AGENTS.md says that signoff runs finish
• 8. Sign off: just signoff passes: Signed off on eb6a57c85c9fc025b5adfeeeba0e46126e7d33ca

Reviewed with sol-loop: 4 rounds with gpt-6.1-sol on xhigh. Sol verdict: approved. Round 2 found S1, a blocker: the severity check skipped headings, **Limit** bullets, and the line that opens fenced code. Round 3 found S2, a should: a sibling ## Severity assessment failed finish. Each fix has a case that fails on the code before it. Round 4 approved.

Closes #412
Closes #413

by Opus-5.5 via Claude Code

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

Run configuration

• Source mode: Auto
• Repository: Unavailable
• Branch: Unavailable
• File: No configuration file from this repository applied
• Commit: Unavailable
• Layers: Repository UI
• Review profile: ASSERTIVE
• Plan: Advanced
• Run ID: 5f72db50-9e08-4ba2-b234-4b564c5132ab

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

_{Comment @coderabbitai help to get the list of available commands.}