review-fixes
Close out an EXISTING code-review findings report — one subagent per finding, each with a regression test — then regenerate EN/zh/ja/ko docs, run all gates, and verify no stale or unexported symbols remain. Trigger on "/review-fixes", "fix all N review findings", "close the review findings", or when the user hands over a findings report to remediate. Distinct from review-remediate, which produces the review itself first — this skill starts from a report that already exists.
- License 许可
- MIT
- 最近更新
- 2026-11-01
- Stars
- ★ 2
- Forks
- 1
01 前言 Front Matter
- name
- review-fixes
- version
- 1.0.0
- description
- Close out an EXISTING code-review findings report — one subagent per finding, each with a regression test — then regenerate EN/zh/ja/ko docs, run all gates, and verify no stale or unexported symbols remain. Trigger on "/review-fixes", "fix all N review findings", "close the review findings", or when the user hands over a findings report to remediate. Distinct from review-remediate, which produces the review itself first — this skill starts from a report that already exists.
03 SKILL.md 正文
Review Fixes — fan out, fix, verify
Remediate every finding in an existing review report via dedicated subagents, then prove the monorepo is clean. Run end-to-end without stopping to ask how to proceed.
1. Read the findings report
- Use the report path given as the argument; otherwise take the newest file in
docs/code-review/(git-ignored working space). - Validate it is non-empty and parseable into discrete findings. If it is missing, empty, or has zero actionable findings, stop and report the gap — do not invent findings.
- Number every finding (
P0-1,P1-2, … or reuse the report's own ids) and create one task per finding (TaskCreate).
2. One subagent per finding
- Spawn a dedicated subagent per finding (parallel is fine for editing). Each subagent must:
- Fix the root cause, not a band-aid, following repo conventions (Reuse-First,
@anvilkit/uiprimitives, Biome tabs, no new deps without approval). - Add a regression test that fails before the fix and passes after. Test-writing gotcha: the react-library vitest preset has
globals: false, so multi-render.tsxtests need an explicitafterEach(cleanup). - Run only its own package's
typecheck+test— never a workspace-wide or parallel build (rslib buildCache SIGABRT and collab submodule races make concurrent builds unreliable; the coordinator builds later, serially via Turbo). - Return: files touched (flagging which are inside submodules), test added, and every symbol it renamed, removed, or newly exported.
- Fix the root cause, not a band-aid, following repo conventions (Reuse-First,
- Verify each subagent actually finished. A subagent that dies mid-run (session limit) leaves stale type references behind. If a report is missing, truncated, or its claimed edits are absent from disk, re-dispatch that finding — do not assume it landed.
3. Regenerate docs and i18n
- If any fix changed a component/plugin public API, run the docs generation for
apps/docs/content/docs/{components,api,templates}(generated content is committed). - Regenerate zh/ja/ko bodies via the central translation store at
apps/docs/i18n/readmes/and runcheck:i18n. Theapireference is EN-only by design — do not translate it. - Escape YAML frontmatter values starting with
@and JSX-like strings in MDX tables, then confirmpnpm docs:buildparses.
4. Run all gates (coordinator, serial)
- Full gate for every touched package:
pnpm typecheck,pnpm lint,pnpm test,pnpm build; addpnpm madgeandpnpm publintfor package-level work. - Rebuild any package that ships from
dist/before declaring a runtime-facing fix done. Note: a barerslib buildof core wipesstyles.css— use the fullpnpm build. - Separate pre-existing failures (verify against clean
HEAD, report them) from regressions you introduced (fix them). Never weaken a gate to get green.
5. Stale / unexported symbol sweep
- Collect the union of symbols the subagents renamed, removed, or added, then:
rgthe workspace (excludingnode_modules) for the OLD names — zero hits allowed outside changelogs/reports.- Confirm every newly referenced symbol is actually exported from its package entry (
rgthe package's public entrypoints; workspacepnpm typecheckmust be green). - If fixes touched path aliases, grep the built
dist/*.d.tsfor leftover@/specifiers (rslib rewrites aliases in.jsbut not.d.ts).
6. Report
- Summarize per finding id: fix, regression test, gate results; list docs/i18n regenerated; state the stale-symbol sweep result; document pre-existing failures separately.
- Leave everything unstaged and flag which modified files live inside submodules. Never commit — this repo is user-committed only.
04 同仓库更多 Skills
react-doctor
Use when finishing a feature, fixing a bug, before committing React code, or when the user types `/doctor`, asks to scan, triage, or clean up React diagnostics. Covers lint, accessibility, bundle size, architecture. Includes a regression check and a full local-triage workflow that fetches the canonical playbook.
review-remediate
Use for the review-then-fix-everything loop on a package or diff — deliver a grounded review report immediately (no scoping questions, no plan-approval), then remediate every finding as tracked numbered items, gating each batch on typecheck/lint/test/build plus an independent Codex pass until clean. Trigger on "review and fix <package>", "remediate the review", "review-fix-verify", or when the user wants a deep review followed by autonomous remediation. Distinct from gstack `review` (pre-landing PR review) and `code-review` (diff only) — this reviews from scratch AND fixes to green.
release-prep
Pre-release verification + go/no-go checklist for publishing @anvilkit packages. Detects the release surface (main workspace fixed-group vs components submodule), runs changeset status and the check:all release gates, enforces this repo's publish gotchas (version.ts drift, ui peer ranges, never-publish canvas-templates, bundle/api-snapshot), then hands the exact publish command to the user. Use when asked to "prep a release", "check release readiness", "/release-prep", or before cutting a version. Does NOT commit or publish.
react-doctor
Use when finishing a feature, fixing a bug, before committing React code, or when the user types `/doctor`, asks to scan, triage, or clean up React diagnostics. Covers lint, accessibility, bundle size, architecture. Includes a regression check and a full local-triage workflow that fetches the canonical playbook.
pre-refactor
暂无简介
phase-execute
Autonomous phase runner for numbered plans (docs/plans/*, docs/prd/*): scoped gates per task, full gate only at phase boundaries, checkpoint to .claude/state/phase-run.json after every task so a run resumes at the exact next task, edit-guard.mjs against concurrent writers, explicit E2E servers, flaky quarantine, deliverables report at phase end. Use for "execute phase X of <plan>", "drive the next phase", "phase-execute <plan>", "resume the phase", or "dry-run phase X".