claws-skill-graphlisted

# Skill Graph — the Claws cognition surface as a connected map ## What this is `graph.json` is the machine-readable adjacency graph of every SKILL.md under `.claude/skills/`, generated by `skill-cognition-legion-v2`. It has **37 nodes** across **7 families** and **370 edges**. **Edge types:** `reference` (one skill cites another by name in its body), `chain` (Skill-tool invocation or named follow-up step), `memory` (family co-membership). **Families:** `claws-core` (7), `claws-memory` (6), `claws-runtime` (7), `titans` (13), `class-infra` (1), `meta-cognition` (1), `standalone` (2). **Top hubs:** `claws-do`, `claws-memory-engine`, `mcp-cognition-engine`, `claws-auto-engine`, `claws-prompt-brain`. ## When to invoke - Deciding which downstream skill to invoke from inside another skill. - Auditing the cognition surface for missing nodes, broken chains, or family drift. - Planning a new skill and finding the closest existing cluster to link from. - `/claws-do` classifying a task — graph provides chain edges for next-step candidates. ## When NOT to invoke - Straightforward task dispatch — use `/claws-do` directly. - Runtime state inspection — use `/claws-stats` or `claws_list`. - Mission text authoring — use `/claws-prompt-brain`. - Memory lookup — use `/claws-recall`. ## Steps 1. Open `graph.json` to browse nodes and edges. 2. Filter edges by `from` = your skill slug to see which skills it chains to. 3. For a new skill: find 2–3 skills in the nearest family, add valid