claude-git-hooks/templates/presets/ai/ANALYSIS_PROMPT.md

Git hooks with Claude CLI for code analysis and automatic commit messages

You are analyzing a **{{PRESET_NAME}}** project with the following technology stack:

**Tech Stack:** {{TECH_STACK}}

**Analyzing files matching:** {{FILE_EXTENSIONS}}

## Your Task

Perform a comprehensive code quality analysis focusing on these areas:

{{FOCUS_AREAS}}

## Analysis Guidelines

1. **Claude API Best Practices**:
   - Proper use of Claude API endpoints
   - Correct model selection (haiku/sonnet/opus)
   - Token usage optimization
   - Error handling for API failures
   - Rate limiting considerations
   - Timeout handling

2. **Prompt Engineering**:
   - Clear and specific instructions
   - Well-structured prompts
   - Appropriate context inclusion
   - Effective use of examples
   - Proper output format specifications
   - Token-efficient prompting

3. **CLI User Experience**:
   - Clear, helpful error messages
   - Appropriate use of colors/formatting
   - Progress indicators for long operations
   - Helpful usage instructions
   - Graceful error recovery
   - Cross-platform compatibility

4. **Git Operations**:
   - Safe git command execution
   - Proper error handling
   - Repository state validation
   - Avoiding destructive operations
   - Cross-platform path handling

5. **Security**:
   - API key protection (never log/expose)
   - Secure credential storage
   - Input validation
   - Avoiding command injection
   - Sensitive data handling in prompts

6. **Code Quality**:
   - Proper error handling
   - Comprehensive logging
   - Modular, reusable functions
   - Clear documentation
   - Maintainable code structure

## Specific Checks

### Claude API Usage
✅ Using appropriate model for task (haiku for simple, sonnet for complex)
✅ Implementing proper error handling (network, API, rate limit)
✅ Token usage calculated and optimized
✅ Prompts are well-structured and clear
✅ Output parsing is robust
✅ API keys never logged or exposed

### Git Operations
✅ Using proper git commands
✅ Handling errors gracefully
✅ Cross-platform path compatibility
✅ Avoiding destructive operations without confirmation
✅ Proper handling of special characters in filenames

### Template Quality
✅ Clear instructions for Claude
✅ Well-defined output format
✅ Appropriate context provided
✅ Token-efficient design
✅ Placeholders used correctly

## Output Format

Respond with a valid JSON following the structured JSON format:

```json
{
  "QUALITY_GATE": "PASSED|FAILED",
  "approved": true|false,
  "metrics": {
    "reliability": "A|B|C|D|E",
    "security": "A|B|C|D|E",
    "maintainability": "A|B|C|D|E",
    "coverage": 0-100,
    "duplications": 0-100,
    "complexity": "number"
  },
  "issues": {
    "blocker": 0,
    "critical": 0,
    "major": 0,
    "minor": 0,
    "info": 0
  },
  "details": [
    {
      "severity": "BLOCKER|CRITICAL|MAJOR|MINOR|INFO",
      "type": "BUG|VULNERABILITY|CODE_SMELL|PROMPT_QUALITY",
      "file": "path/to/file",
      "line": 123,
      "message": "Clear description of the issue"
    }
  ],
  "securityHotspots": 0,
  "blockingIssues": ["List of critical issues that must be fixed"]
}
```

## Analysis Rules

- **Block commit** if:
  - Exposed API keys or secrets
  - Security vulnerabilities
  - Critical bugs in git operations
  - Destructive operations without safeguards

- **Pass** if: Only minor issues, suggestions, or no issues

- Be constructive about prompt quality - suggest improvements
- Focus on safety and user experience
- Provide actionable, specific feedback with line numbers