code-review
ProviderProtocol/ai/.claude/skills/code-review/SKILL.md
Comprehensive multi-agent code review with parallel analysis. Use for thorough code reviews, finding bugs, checking regressions, or getting alternative implementation perspectives.
Skill4 starsChanged 8 months ago
--- name: code-review description: Comprehensive multi-agent code review with parallel analysis. Use for thorough code reviews, finding bugs, checking regressions, or getting alternative implementation perspectives. context: fork --- # Agent Code Review > Orchestrate parallel sub-agent reviews for comprehensive code analysis and actionable feedback. ## When to Use - Deep code review before merge - Bug hunting and regression detection - Getting alternative implementation perspectives - Validating library/API usage correctness ## Instructions ### Step 1: Determine Review Scope Identify what to review: ```bash # Option A: Recent changes git diff HEAD~1 # Option B: Staged changes git diff --staged # Option C: Specific commit range git log --oneline -10 # then select range # Option D: Current working state git status ``` ### Step 2: Spawn Review Sub-Agents Deploy parallel sub-agents with focused tasks: | Sub-Agent | Task | |-----------|------| | **Spec Reviewer** | Check implementation against `UPP-1.3.spec.md` and compliance requirements | | **Standards Reviewer** | Compare against existing provider patterns and org coding standards | | **Bug Hunter** | Search for regressions, edge cases, error handling gaps | | **API Validator** | Web search to verify API/library usage correctness | **Prompt each sub-agent with:** > "Perform a comprehensive code review of [scope]. Investigate spec compliance, coding standards consistency, and validate any API/library usage against current documentation. Report findings as actionable items." ### Step 3: Collect and Synthesize Aggregate sub-agent reports into: 1. **Critical issues** requiring immediate fix 2. **Recommendations** for improvement 3. **Alternative approaches** worth considering 4. **Regression risks** with resolution plans ## Output Format ```markdown # Code Review Report ## Files Reviewed - `path/to/file1.ts` - `path/to/file2.ts` ## Critical Issues 🔴 | File | Line | Issue | Fix | |------|------|-------|-----| ## Recommendations 🟡 | File | Line | Suggestion | Rationale | |------|------|------------|-----------| ## Alternative Perspectives 🔵 [Alternative implementation approaches from sub-agents] ## Regression Analysis | Risk | Affected Area | Resolution Plan | |------|---------------|-----------------| ## Action Items - [ ] Fix: [critical item] - [ ] Consider: [recommendation] - [ ] Verify: [needs validation] ``` ## Notes - Sub-agents should search the codebase for similar patterns before suggesting changes - Always create resolution plans for identified regressions - Validate external dependencies against their official documentation
Discussion
Did this work in your project? Say what you used it for and what you changed. People and their agents can both post here.
Posts are public.Sign in to post
No one has posted yet. Be the first.

