Orchestrate specialized review agents for code quality assessment. Use when user mentions レビューして, コードレビュー, 品質チェック, code review, quality check. Do NOT use for quick PR screening (use /preview instead).
From shipnpx claudepluginhub thkt/dotclaude --plugin toolkitThis skill is limited to using the following tools:
references/pre-flight.mdProvides UI/UX resources: 50+ styles, color palettes, font pairings, guidelines, charts for web/mobile across React, Next.js, Vue, Svelte, Tailwind, React Native, Flutter. Aids planning, building, reviewing interfaces.
Fetches up-to-date documentation from Context7 for libraries and frameworks like React, Next.js, Prisma. Use for setup questions, API references, and code examples.
Calculates TAM/SAM/SOM using top-down, bottom-up, and value theory methodologies for market sizing, revenue estimation, and startup validation.
Orchestrate specialized review agents with confidence-based filtering. The
finding schema requires file:line on every finding — entries without evidence
are structurally invalid.
| Excuse | Counter |
|---|---|
| "This is a false positive" | Verify with evidence-verifier before dismissing. Intuition ≠ evidence |
| "This pattern is intentional" | No // intentional: marker = not intentional |
| "Low severity, skip it" | Low severity × high frequency = high risk. Count occurrences |
| "The code works fine" | Working ≠ correct. Audit reviews quality, not functionality |
| "This is third-party code" | If it's in your repo, it's your responsibility |
$1 (optional)$1 is empty → select focus via AskUserQuestion, then review
staged/modified files| Question | Options |
|---|---|
| Focus | security / performance / readability / all |
Start with Pre-flight (see below). Save snapshot before displaying any results to user.
| Step | Action |
|---|---|
| 1 | Pre-flight (tests + hook findings) |
| 2 | File routing: classify target files → assign to relevant reviewers |
| 3 | Spawn sub-reviewers via Task (background, max 10 parallel) |
| 4 | Spawn challenger + verifier (wait for reviewers) |
| 5 | Spawn integrator (wait for challenger + verifier) |
| 6 | Leader receives final Markdown from integrator |
| 7 | Save snapshot |
| 8 | Display delta + report |
Leader classifies each target file by path and assigns to relevant reviewers only:
| File Pattern | Sub-reviewers (subagent_type) |
|---|---|
*.sh | security-reviewer, silent-failure-reviewer, code-quality-reviewer, |
| duplication-reviewer, reuse-reviewer, efficiency-reviewer, | |
| operational-readiness-reviewer | |
*.ts, *.tsx, *.js | security-reviewer, silent-failure-reviewer, type-safety-reviewer, |
| code-quality-reviewer, duplication-reviewer, reuse-reviewer, | |
| efficiency-reviewer, design-pattern-reviewer, testability-reviewer, | |
| performance-reviewer, operational-readiness-reviewer | |
*.md (agent defs) | prompt-reviewer, document-reviewer |
*.md (skills/docs) | prompt-reviewer, document-reviewer |
*.md (rules) | prompt-reviewer, document-reviewer |
*.yaml, *.json | type-design-reviewer, document-reviewer |
*.css, *.html | accessibility-reviewer, progressive-enhancer, performance-reviewer, |
| duplication-reviewer | |
test.*, *.test.* | test-coverage-reviewer, testability-reviewer |
| Other | code-quality-reviewer, duplication-reviewer, reuse-reviewer, |
| efficiency-reviewer, document-reviewer |
Classification by path: agents/**/*.md → agent defs, skills/*/SKILL.md or
docs/**/*.md → skills/docs, rules/**/*.md → rules, other *.md →
skills/docs (default).
root-cause-reviewer is not in this table — it runs sequentially after code-quality-reviewer completes (see Sequential Dependencies). Leader spawns it with the same file list + CQ findings as input.
Each sub-reviewer is spawned directly via Task:
security-reviewer)| Role | subagent_type | Purpose |
|---|---|---|
| challenger | devils-advocate-audit | Challenge findings (reduce FP) |
| verifier | evidence-verifier | Verify findings (positive evidence) |
| integrator | progressive-integrator | Reconcile into root causes |
| Reviewer | Depends On | Reason |
|---|---|---|
| root-cause | code-quality | Needs CQ findings as input |
| challenger | All reviewers | Needs all findings |
| verifier | All reviewers | Needs all findings |
| integrator | challenger + verifier | Needs both perspectives |
Agents are standalone. Leader collects via Task completion; spawns via Task prompt.
| Error | Recovery |
|---|---|
| No files to audit | Return "No files to audit" |
| Reviewer stall | 120s timeout; proceed without |
| Malformed output | Skip reviewer, log warning, proceed with valid reviewers |
| Dependency stall | Skip dependent (e.g., root-cause if CQ failed) |
| Max parallel >10 | Batch in groups of 10 |
| Challenger stall | 120s timeout; proceed with verifier only |
| Verifier stall | 120s timeout; proceed with challenger only |
| Integrator stall | 120s timeout; Leader integrates manually |
Read references/pre-flight.md for the full
procedure: detect task runner → find test script → run tests → convert hook
output to PF-{seq} findings.
Session ID: ${CLAUDE_SESSION_ID}
SNAPSHOT="$HOME/.claude/workspace/history/audit-$(date -u +%Y-%m-%d-%H%M%S).yaml"
| Template | Purpose |
|---|---|
| @../templates/audit/output.md | Output format with delta |
| @../templates/audit/snapshot.yaml | Snapshot schema |
| Check | Required |
|---|---|
| Reviewers completed? | Yes |
| Challenger validated? | Yes |
| Verifier verified? | Yes |
| Integrator produced Markdown? | Yes |
| Snapshot saved? | Yes |
| Delta displayed? | Yes |