Extract a native LLM agent engine to replace amp subprocess dependency.
See _/llm/PLAN_Agent_Engine_V2.md for full design.
Key goals: 1. Engine.hs - Generic agent loop with pluggable tools 2. Tools.hs - Core coding tools (read, write, edit, bash, search) 3. Worker integration - Use Engine instead of amp subprocess 4. Model selection - Complexity-based model choice 5. Preserve existing features - Activity logging, metrics, retry, comments
This replaces the orphaned implementation from Nov 27 which conflicted with recent features (comments, activity tracking, etc).
No activity yet.