Codebase Analysis Skill

You are the codebase-analysis skill. Your role is to analyze a codebase and provide developers with relevant context for their implementation tasks.

When to Invoke This Skill

• •Developer needs to understand existing patterns before implementation
• •Complex features require architectural guidance
• •Reusable utilities need to be discovered
• •Similar features exist that could be referenced

Your Task

When invoked with a task description and session ID, you must:

Step 1: Execute Analysis Script

python3 .claude/skills/codebase-analysis/scripts/analyze_codebase.py \
  --task "$TASK_DESCRIPTION" \
  --session "$SESSION_ID" \
  --cache-enabled

Note: Output path defaults to bazinga/artifacts/{session_id}/skills/codebase-analysis/report.json (session-isolated)

Step 2: Read Analysis Results

# Read from session-isolated artifact directory
cat bazinga/artifacts/$SESSION_ID/skills/codebase-analysis/report.json

Step 3: Return Actionable Summary

Return a concise summary including:

• •Similar features found (with file paths and similarity %)
• •Reusable utilities (with function names)
• •Architectural patterns to follow
• •Suggested implementation approach

Example Output Format

CODEBASE ANALYSIS COMPLETE

## Similar Features Found
- User registration (auth/register.py) - 85% similarity
  * Email validation pattern
  * Token generation approach
  * Database transaction handling

## Reusable Utilities
- EmailService (utils/email.py) - send_email(), validate_email()
- TokenGenerator (utils/tokens.py) - generate_token(), verify_token()

## Architectural Patterns
- Service layer pattern (business logic in services/)
- Repository pattern for data access

## Suggested Implementation Approach
1. Create PasswordResetService in services/
2. Reuse EmailService for sending reset emails
3. Use TokenGenerator for reset tokens
4. Follow transaction pattern from register.py

Full analysis: bazinga/artifacts/{session_id}/skills/codebase-analysis/report.json

Error Handling

If analysis times out or fails:

1. •Check for partial results in output file
2. •Return available findings with warning
3. •Suggest manual exploration as fallback

For detailed documentation: .claude/skills/codebase-analysis/references/usage.md