[nsheaps]

Phase 1 plan, M0 PR 6 (REQ-ENG-006). Builds on #367, which formatted the repo and added .prettierignore.

Changes

• mise run lint checks formatting. A new lint:format task runs prettier --check ., and lint depends on it, so the lint job fails on unformatted files.
• .github/workflows/format.yml autofixes PRs from this repo.
  • It checks out the PR head with the automation App token (checkout-as-app, SHA-pinned), runs mise run format, and pushes any changes as style: apply prettier formatting via scripts/ci/commit-format-fixes.sh.
  • The push uses the App token rather than GITHUB_TOKEN, so CI re-runs on the autofix commit. It never adds [skip ci].
  • Concurrency is per PR.
• Fork PRs can't receive pushes. A separate job warns the author to run mise run format, and their lint job still fails.
• Docs:
  • REQ-ENG-006 moves to partial; eslint --fix is still missing.
  • Open question 2 is decided: Prettier is the formatter of record.
  • CLAUDE.md and CONTRIBUTING.md no longer claim that bun run lint runs Prettier, and they list mise run lint:format and mise run format.

Test plan

• Locally, a deliberately misformatted file made mise run lint:format fail, and it passed once the file was removed.
• mise run lint:workflows passes (actionlint, shellcheck, single-command run: steps).
• mise run check passes.
• Plan acceptance, on this PR: push a commit with a formatting violation. Lint should fail, Format should push an autofix commit, and the PR should go green.

🤖 Generated with Claude Code

https://claude.ai/code/session_01GA968hkqa7CsMeYAhzV4Dz

---

Generated by Claude Code

[github-actions]

Release Version Check

🟢 PATCH version bump

	Version
Current	0.7.42
Next	0.7.43

This version will be released automatically when this PR is merged to main.

Changelog preview

0.7.43 (2026-10-08)

Tests

• deliberately misformat pr-title.ts to exercise the format autofix (cfef16b)

CI/CD

• add format check and pr autofix (44d4af6)
• commit untracked files in format autofix; clarify contributing note (c68427a)

Styles

• apply prettier formatting (17aabf1)

[github-actions]

Preview Deployment

The web app for this PR has been deployed:

Open Preview

Use this to verify the app works correctly, especially for dependency updates.

[nsheaps]

Autofix acceptance check: passed.

1. cfef16b deliberately misformatted scripts/ci/pr-title.ts.
2. The Format workflow then pushed 17aabf1 (style: apply prettier formatting, authored by automation-nsheaps[bot]) as a new commit. 17aabf1 restores the file, so the PR has no net change to pr-title.ts.
3. CI is now running on 17aabf1.

deploy-preview failed on cfef16b, but not because of this PR. Its git push to gh-pages was rejected because the remote had moved on. 44d4af6 and cfef16b were pushed about a minute apart, and preview-deploy.yml has no concurrency group, so the two preview runs raced on gh-pages. That commit is now superseded, so I'm not re-running it; the preview runs again on 17aabf1. Autofix commits will make back-to-back pushes more common. Serialising gh-pages writers with a shared concurrency group belongs in a separate CI PR, not this one.

---

Generated by Claude Code

[henry-nsheaps]

### ✅ Post-merge re-review — no new content since prior APPROVE

• ✅ PR is already merged (merge commit 11dc06b) at head SHA c68427a.
• ✅ My prior APPROVED review (PRR_kwDORdvHSc8AAAABRP8tnQ) was against this exact SHA; the diff hasn't changed since.
• ✅ All four previous inline threads are resolved (3× P2 addressed by author, 1 info note confirmed).
• ✅ No new findings in a fresh pass over format.yml, commit-format-fixes.sh, .mise.toml, and the REQ-ENG-006 doc edits.

Click to expand for full details

Why this is a comment, not an approve

A fresh APPROVE on a merged PR adds noise; my previous APPROVE already stands against the merged SHA. Posting as COMMENT here just records that this re-review run found nothing new.

How I re-checked

• gh pr view 368 confirmed state: MERGED, headRefOid: c68427a…, mergeCommit.oid: 11dc06b….
• gh pr diff 368 matched the diff I reviewed previously — zero new commits since c68427a1.
• Verified all four review threads (PRRT_kwDORdvHSc6qOsmn, …sw6, …s8H, …tK8) are already isResolved: true.
• Minimized my two prior review bodies as OUTDATED so the thread view stays clean. The APPROVE verdict for c68427a remains in the review history.

Nothing re-opened

• .github/workflows/format.yml: still pull_request (not _target), App token gated by head.repo.full_name == github.repository, SHA-pinned checkout-as-app.
• scripts/ci/commit-format-fixes.sh: untracked-files guard + git add --all from c68427a present.
• Docs: REQ-ENG-006 is partial / documented / current with eslint --fix called out as the remaining gap; CLAUDE.md and CONTRIBUTING.md no longer claim bun run lint runs Prettier; Open Question 2 is marked decided.

Follow-ups: none for this PR.

• (Tracked separately, not for this PR.) eslint --fix wiring once REQ-ENG-007 extends ESLint coverage — already captured in REQ-ENG-006 "Gap".

References

• Workflow run: https://github.com/nsheaps/.ai-agent-henry/actions/runs/37738578193
• Prior APPROVE: PRR_kwDORdvHSc8AAAABRP8tnQ (now minimized as OUTDATED — verdict still applies to the merged SHA)
• GitHub docs — Minimizing conversations