add-artefact-graphlisted

# Artefact Graph — Querying It Owns HOW the artefact graph is queried: which verb answers which question, the two interfaces, the case of an agent that can reach neither, and what the answer does not cover. WHAT a given command does with the answer belongs to that command. **A command states its question and points here.** It does not name a verb, because a named verb is what holds an agent to one query when the question needed two. **The graph is emitted by `node scripts/build.js` into `framwork/.codeadd/artefact-graph.json`, and it covers `workbench/` as well as `framwork/.codeadd/`.** It is rebuilt from scratch on every build and is gitignored, so it holds no history — `history` is the one verb that reads something else. ## When to Use - About to ask what an artefact relates to, depends on, or would break. - Grading the blast radius of a change before planning or executing it. - Checking whether an artefact still has a consumer. ## When NOT to Use - Looking for TEXT. Grep is the right tool for finding a phrase, and this skill never says otherwise. - Asking what SHIPPED and was later dropped. That is the delivery index, reached through `history`. --- ## The Rule ``` IF THE QUESTION IS "WHAT DOES THIS ARTEFACT RELATE TO": ⛔ DO NOT USE: Grep to reconstruct a relationship from prose ✅ DO: Ask the graph ✅ DO: Report the answer as NOT VERIFIED when the graph does not answer ``` ⛔ **The rule is "ask the graph", never "use the MCP instead of grep".** Those are dif