What problem does it solve? Mature projects accumulate settled vocabulary and decisions scattered across CLAUDE.md, README files, decision logs, and transcripts, with no single authoritative glossary. This Skill performs a one-off adoption pass that harvests that existing knowledge into a structured DOMAIN.md and docs/decisions/ records without inventing new terminology. ## Core Features & Use Cases - Guided five-step pass: Explore sources, propose a candidate model for review, ask one batched round of questions about conflicts, write the files, and report findings — nothing is written before operator review. - Multi-source harvesting: Mines orientation files, decision logs, folder names, live registries, prior session transcripts, and an operator narration to derive terms rather than invent them. - Decision log conversion: Sorts existing decision tables through a gate (hard to reverse, real alternatives), dates records from git history, and renumbers them chronologically. - Use Case: A two-year-old project with a stale CLAUDE.md and a DECISIONS.md table gets a reviewed DOMAIN.md of 14 terms, six numbered decision records dated from git history, and a report flagging where documentation disagrees with what actually ships. ## Quick Start Ask the AI to adopt or seed a domain model for this project using the domain-adopt skill.
