learning-loop
SKILL.md
Learning Loop
A proactive learning engine. Your agent reflects, journals, tracks decisions, extracts lessons, and gets smarter every day — without being told to.
What Makes This Different
Most self-improvement skills are reactive — they only learn when corrected. Learning Loop is proactive:
- Journals daily — synthesizes what happened, not just what went wrong
- Tracks decisions with categories — analyze accuracy by type over time
- Generalizes lessons — asks "where else does this apply?" before saving
- Manages lesson lifecycle — ACTIVE → VALIDATED → CHALLENGED → ARCHIVED
- Tiers memory automatically — hot/warm/cold based on recency and usage
- Staggered review chain — nightly → monthly → quarterly → year-end
- Correction tracking — counts repetitions, promotes to rules after 3x
- Quiet day protocol — idle time becomes maintenance time