v1.2 — Quick Learning & Self-Improving Process

What's New

Quick Learning Skill

• Runs automatically as a background subagent at every session break
• Analyzes reasoning patterns, not specific technical decisions
• Writes transferable insights to shared knowledge base (reasoning-patterns.md)
• Zero main context token cost — runs in isolated subagent context
• Also available manually: /quick-learning

Rewritten README

• Properly explains methodology principles and philosophy
• Full pipeline documentation with all 6 stages
• Validation matrix showing what each validator catches
• Self-improvement loop explanation (quick-learning vs retrospective)

Integration Points

• feature-execution: Step a0 before session end protocol
• do-task: Quick Learning before session handoff
• methodology: documented in skills table and commands reference
• CLAUDE.md: registered as user-invocable skill

Full Changelog

See README for complete v1.1 changelog (folded under details).