distill

Distill session episodes into persistent memory patterns

by c-daly|Open Source

Usage

/distill           # Distill episodes at current scope
/distill show      # Show current memory
/distill episodes  # Show pending episodes

How It Works

Extract patterns from .context/EPISODES.md learnings
Classify as pattern, pitfall, preference, or approach
Match against existing .context/MEMORY.md patterns
Reinforce matches (increases confidence), add new, decay old, prune low

CLI

python3 ~/.claude/plugins/agent-swarm/context/memory.py distill .   # Distill
python3 ~/.claude/plugins/agent-swarm/context/memory.py show .      # Show
python3 ~/.claude/plugins/agent-swarm/context/memory.py episodes .  # Pending

Confidence

Reinforcement: each observation increases confidence
Decay: patterns not seen in 30+ days lose confidence
Pruning: below 0.2 removed

Logging

Agents log learnings with: LEARNING: [description] Captured by post-task hooks -> EPISODES.md.

distill - AI Agent Skill for Claude Code & Cursor | Agent Skills