absorblisted

# /absorb — Session Knowledge Distillation > "What did this session teach us that no file in the repo can?" Extract project understanding from the current session and persist it to memory. Not code patterns — those live in the code. This captures **intent, direction, vision, domain knowledge, and decisions** that would otherwise evaporate when context clears. ## Memory Location By default, this skill writes to `~/.claude/memory/`. To use a different location, set the `CLAUDE_MEMORY_DIR` environment variable: ```bash export CLAUDE_MEMORY_DIR="$HOME/my-claude-memory" ``` Throughout this skill, `$MEMORY` refers to that directory (`${CLAUDE_MEMORY_DIR:-$HOME/.claude/memory}`). ## When to Use - End of a deep session where you learned things about a project's purpose or direction - After domain modeling discussions (data schemas, taxonomies, business rules) - After architectural decisions that have a *why* behind them - When the user shared context about stakeholders, timelines, or vision - Before `/compact` or session end to preserve understanding ## Arguments ``` /absorb — Full session distillation /absorb @project:name — Scope to a specific project /absorb @dry — Show what would be saved without writing ``` --- ## Phase 1: NIGREDO — Extract Raw Knowledge Scan the entire conversation for knowledge that is NOT derivable from code, git history, or existing docs. ### 1.1 Classify Findings For each insight, classify into exactly on