@morodomi/ait3
Version:
AIT³ Development Platform - AI + Ticket + Test + Tool driven development methodology
80 lines (64 loc) • 2.01 kB
JavaScript
/**
* Multi-agent code review command guide template
* Provides structured approach for comprehensive code reviews using Claude and Gemini
*/
export const reviewTemplate = `# Multi-Agent Code Review
Perform comprehensive code reviews using multiple AI perspectives.
## Review Process
### Step 1: Correctness Review (Claude)
Review the current changes for correctness:
- Logic errors and bugs
- Edge case handling
- Adherence to requirements
- Code consistency
### Step 2: Performance Review (Gemini)
Switch to terminal and run:
\`\`\`bash
gemini -p "@src/ @tests/ Review these changes for performance:
Focus on:
- Algorithm efficiency
- Database query optimization
- Memory usage patterns
- Potential bottlenecks
- Async/await optimization"
\`\`\`
### Step 3: Security Review (Claude)
Review the changes for security issues:
- Input validation
- Authentication/authorization
- SQL injection risks
- XSS vulnerabilities
- Sensitive data exposure
- Dependency vulnerabilities
### Step 4: Synthesize Results
Create a summary that includes:
1. Critical issues from all reviews
2. Conflicting recommendations (if any)
3. Priority-ordered action items
4. Overall assessment
### Step 5: Document Decision
After human review, document the decision:
\`\`\`bash
git add .
git commit -m "review(#TICKET): implement review feedback
Correctness: [summary of fixes]
Performance: [optimizations made]
Security: [vulnerabilities addressed]
Human decision: [rationale for choices made]"
\`\`\`
## Quick Commands
For specific review types:
\`\`\`bash
# Performance only
gemini -p "@src/ Review this code for performance bottlenecks"
# Security only
# In Claude: "Review this code for security vulnerabilities"
# Architecture review
gemini -p "@src/ @docs/ Is this architecture scalable and maintainable?"
\`\`\`
## Tips
- Run reviews after GREEN phase, before REFACTOR
- Document conflicting opinions in commit messages
- Keep context focused on specific changes
- Use ticket scope to limit review scope
`;