| name | query |
| description | Put this project's graph to work — answer a grounded question, or draft and advance a deliverable in work/, using the typed notes, their links, and golden context, and file genuinely new conclusions back. Use when the user asks something the project should know, or asks to move a deliverable forward. |
query
Use the graph toward the goal, and file useful results back. Read CLAUDE.md first.
Steps
- Read
PROJECT.md and knowledge/index.md first — don't blind-scan. Pull the relevant notes and
follow their links one hop at a time; the edges are the point. Load knowledge/golden/ —
definitions) as authoritative constraints, and read the knowledge/templates/ types to resolve
names against them so the same concept isn't reinvented.
- Answer the question, or advance a deliverable in
work/, citing the notes you used (link to them).
Treat golden context as hard constraints — flag anything that conflicts with it rather than
overriding it.
- Respect locked work: don't rewrite a
locked: true file without asking. Advance only the unlocked
parts.
- File back: when a genuinely new conclusion or connection emerges, add it as a typed note, link
it into the graph, and update
knowledge/index.md so it compounds for next time.
- If the graph can't answer well, say so plainly and suggest what to ingest next — don't fabricate.
When a follow-up needs the user to choose — which deliverable to advance, which reading of an ambiguous
reference, whether to file a new note — ask with the structured AskUserQuestion prompt (CLAUDE.md
§14), not a free-text question.
Don'ts
Don't answer from memory when the graph should know. Don't overwrite golden or locked files. Don't drop
a frontmatter key you didn't write. Don't make a claim the notes don't support without flagging it.