review-quality（审查质量）

Quality Review (Unified)

Goal: turn “is this change good?” into a repeatable review with a clear merge/production readiness verdict.

This skill intentionally merges three review lenses:

1. Merge readiness (requirements alignment + risk + verification)
2. Code maintainability (Clean Code-style review)
3. Documentation consistency (README/docs vs implementation)

This is the single entry point for Ship Faster reviews. It includes an internal auto-triage:

• Always run the unified review (this skill).
• If React/Next.js performance risk is detected, also run review-react-best-practices and include its findings.
• If UI surface changes are detected, also run a Web Interface Guidelines audit (a11y/focus/forms/motion/content overflow) and include terse file:line findings.

Inputs (recommended)

• BASE_SHA and HEAD_SHA for diff-based reviews
• Optional: PLAN_OR_REQUIREMENTS path (e.g. run_dir/evidence/features/<feature_slug>-plan.md)
• Optional: docs scope (README.md, docs/**, API contracts)
• Optional: run_dir (Ship Faster run directory, if available)

Suggested scope commands

# Baseline: main/master merge-base
BASE_SHA=$(git merge-base HEAD main 2>/dev/null || git merge-base HEAD master)
HEAD_SHA=$(git rev-parse HEAD)

git diff --stat "$BASE_SHA..$HEAD_SHA"
git diff "$BASE_SHA..$HEAD_SHA"

Process

0) Auto-triage (Built-in Router)

Goal: reduce user choice. The reviewer decides which specialized review lens to apply, deterministically, based on the diff.

Collect signals (minimum):

git diff --name-only "$BASE_SHA..$HEAD_SHA"
git diff --stat "$BASE_SHA..$HEAD_SHA"
git diff "$BASE_SHA..$HEAD_SHA"

Routing rules (apply in order):

1. Docs-only change (fast path):

- If every changed file is in docs/** or ends with .md|.txt|.rst - Then run the unified review, but focus on Docs ↔ Code consistency + release risk (skip deep code maintainability unless docs reference code changes).

1. React/Next.js performance-sensitive change:

- If any changed file matches: - **/*.tsx, **/*.jsx - next.config.*, app/**, pages/**, src/app/**, src/pages/** - React/Next entrypoints (layout.tsx, page.tsx, middleware.ts, route.ts, loading.tsx) - Or the diff contains performance-sensitive keywords (spot check via git diff): - use client, Suspense, dynamic(, next/dynamic, next/navigation, React.cache, revalidate, fetch(, headers(), cookies() - Then: run review-react-best-practices and include its output (or link to its artifact) in the final report.

1. UI Web Interface Guidelines audit (a11y/UX rules, terse output):

- If any changed file matches: - UI code: **/*.tsx, **/*.jsx, **/*.vue, **/*.html, **/*.css, **/*.scss - Common UI dirs: app/**, pages/**, src/app/**, src/pages/**, components/**, src/components/** - Then: fetch the latest Web Interface Guidelines and audit only the changed UI files first. - Source: https://raw.githubusercontent.com/vercel-labs/web-interface-guidelines/main/command.md - Fetch method: WebFetch (if available) or curl -fsSL <url> - Output findings in terse file:line format, grouped by file (high signal, low prose). Treat a11y + focus issues as higher severity than cosmetic issues.

Output requirement (at top of your report):

## Triage
- Docs-only: yes|no
- React/Next perf review: yes|no
- UI guidelines audit: yes|no
- Reason: <1-3 bullets based on file paths / patterns>

1) Merge readiness (verdict review)

Use the template: references/code-reviewer.md.

Minimum checks:

• Requirements alignment (acceptance criteria hit, non-goals respected)
• Side effects are gated (deploy/payments/DB writes require explicit approval)
• Error handling is explicit (no silent failures)
• Verification evidence exists (tests/build/typecheck/lint/manual steps)

2) Clean Code maintainability scan

Review the diff (and nearby touched code) across these dimensions:

1. Meaningful naming (avoid data1, tmp, mixed naming for same concept)
2. Small functions / SRP (very long functions, too many params, mixed responsibilities)
3. Duplication (DRY) (copy/paste logic, repeated transformations/validation)
4. Over-engineering (YAGNI) (unused branches, unnecessary abstractions)
5. Magic numbers / strings (hardcoded values without semantic constants)
6. Structural clarity (deep nesting, unreadable one-liners, nested ternaries)
7. Project conventions (imports/order/style consistency)

3) Docs ↔ code consistency scan

Check that docs do not lie:

• Enumerate: README.md, docs/**/*.md, config examples, env keys, API contracts
• For each claim/config/example: locate the authoritative code/config/contracts
• Record mismatches with evidence (doc location + code location)

Output (required)

Produce a structured report:

• quality-review.md
• quality-review.json (optional but recommended)

If you are working inside a Ship Faster run directory, write to:

• run_dir/evidence/quality-review.md
• run_dir/evidence/quality-review.json

If triage selects React/Next performance review and a Ship Faster run_dir is available, also persist:

• run_dir/evidence/react-best-practices-review.md

If triage selects UI guidelines audit and a Ship Faster run_dir is available, also persist:

• run_dir/evidence/ui-guidelines-review.md (terse file:line findings, grouped by file)

Output format (recommended)

## Summary
- Verdict: Ready / With fixes / Not ready
- Scope: BASE_SHA..HEAD_SHA (or file list)

## Triage
- Docs-only: yes|no
- React/Next perf review: yes|no
- UI guidelines audit: yes|no
- Reason: ...

## Strengths
- ...

## Issues
### Critical (Must Fix)
- Location: path:line
  - What
  - Why it matters
  - Minimal fix

### Important (Should Fix)
...

### Minor (Nice to Have)
...

## UI Guidelines (terse, only if audit=yes)
- path:line <finding>
- path:line <finding>