Self-Improving Agent

Implements a continuous 4-phase self-improvement loop: (1) Detection from conversations/tool outputs, (2) Structured Logging into .learnings/ artifacts using LRN/ERR/FEAT schemas with traceable sources, (3) Pattern Analysis with recurrence counting, and (4) Promotion of verified, broadly-applicable learnings into Tenets, LTM, or draft Skill artifacts—preventing repeat mistakes and systematically growing capabilities.