parallel-code-review
Parallel 3-reviewer code review orchestration: launch Security, Business-Logic, and Architecture reviewers simultaneously, aggregate findings by severity, and produce a unified BLOCK/FIX/APPROVE verdi
SKILL.md.claude/skills/parallel-code-review/SKILL.md
GitHubParallel Code Review Skill
Orchestrate three specialized code reviewers (Security, Business Logic, Architecture) in true parallel using the Fan-Out/Fan-In pattern. Each reviewer runs independently with domain-specific focus, then findings are aggregated by severity into a unified BLOCK/FIX/APPROVE verdict.
Instructions
Phase 1: IDENTIFY SCOPE
Goal: Determine changed files and select appropriate agents before dispatching.
Step 1: Read repository CLAUDE.md to load project-specific conventions.
Step 2: List changed files
git diff --name-only HEAD~1
# or for PRs:
gh pr view --json files -q '.files[].path'
Step 3: Select architecture reviewer agent based on dominant language.
Gate: Changed files listed, architecture reviewer agent selected.
Phase 2: DISPATCH PARALLEL REVIEWERS
Goal: Launch all 3 reviewers in a single message for true concurrent execution.
Critical constraint: All three Task calls MUST appear in ONE response.
Reviewer 1 -- Security: OWASP Top 10, authentication, authorization, input validation, secrets exposure.
Reviewer 2 -- Business Logic: Requirements coverage, edge cases, state transitions, data validation, failure modes.
Reviewer 3 -- Architecture: Design patterns, naming, structure, performance, maintainability.
Gate: All 3 Task calls dispatched in a single message. Proceed only when ALL 3 return results.
Phase 3: AGGREGATE
Goal: Merge all findings into a unified severity-classified report.
| Severity | Meaning | Action |
|---|---|---|
| CRITICAL | Security vulnerability, data loss risk | BLOCK merge |
| HIGH | Significant bug, logic error | Fix before merge |
| MEDIUM | Code quality issue, potential problem | Should fix |
| LOW | Minor issue, style preference | Nice to have |
Deduplicate overlapping findings. Keep the highest severity.
Gate: All findings classified, deduplicated, and summarized.
Phase 4: VERDICT
| Condition | Verdict |
|---|---|
| Any CRITICAL findings | BLOCK |
| HIGH findings, no CRITICAL | FIX (fix before merge) |
| Only MEDIUM/LOW findings | APPROVE (with suggestions) |
Gate: Structured report delivered with verdict. Review is complete.
Error Handling
| Error | Solution |
|---|---|
| Reviewer times out | Report partial findings, note blind spots |
| All reviewers fail | Verify file paths, reduce scope, fall back to systematic-code-review |
| Conflicting findings | Keep higher severity, include both perspectives |