FLY-LEARN: Formalize learning capture loops from session logs

Labels: enhancement, fly-methodology, patterns

Description

Currently we capture patterns informally in pattern-catalog.md. We need a systematic process to extract learnings from our work and prevent repeating mistakes.

Current State

• Patterns discovered ad-hoc during sessions
• Manual updates to pattern-catalog.md
• No systematic review process
• Learning often lost in session logs

Desired State

• Automated pattern extraction from session logs
• Weekly pattern review during audits
• Categorized pattern library
• Integration with methodology docs

Success Criteria

• Pattern extraction script/process defined
• Categories established (verification, coordination, error-handling, etc.)
• Weekly review integrated into FLY-AUDIT process
• Pattern-catalog.md auto-updated with discoveries
• Anti-patterns documented (what NOT to do)

Examples to Capture

• Mock data fallbacks causing validation theater
• Infrastructure verification preventing wrong gameplans
• Multi-agent coordination patterns
• Evidence requirements preventing false claims

Estimated: 4 hours Priority: Medium (supports continuous improvement)

Implementation Notes

Pattern Categories

• Verification Patterns: How to validate work properly
• Coordination Patterns: Multi-agent collaboration
• Error Handling: Graceful degradation vs honest failure
• Anti-Patterns: What NOT to do (validation theater, etc.)

Integration Points

• Weekly FLY-AUDIT process
• Session log analysis
• Pattern-catalog.md maintenance
• Methodology documentation updates

Technical Approach

1. Create pattern extraction script for session logs
2. Define pattern taxonomy and templates
3. Integrate with existing audit workflow
4. Build pattern library with search/filter capabilities