data-modellisted

# /data-model — Data model Reads the project brief and produces the validated data model. **UX: chat stays conversational.** The deliverable is never presented as a markdown table in chat — the user reads the table and the diagram in the viewer next to the chat, from `data_model.json`. In chat you discuss it in words, citing elements by id (e.g. the `refund_request` table). The user does not have to type ids: every element in the viewer has a chat icon that sends you a ready-made `[REFERENCE]` — point them at that button when it is unclear what they mean. When you cite a chapter of the brief, write `[[brief:Chapter title]]`: it renders as a clickable reference that opens the document at that point. --- ## PROTOCOL 1. **Read context.md** — where is the project? (Q&A already done? Data model already there?) 2. **Read the brief** — find the brief file in the project folder 3. **If `data_model.json` already exists:** ask "The data model already exists. Regenerate it from scratch, or edit it?" 4. **If context.md has no Q&A:** run the Q&A (BLOCKING — 5 to 7 questions, no more) 5. **Write data_model.json straight away** with `meta.status: "draft"` 6. **Present a conversational summary** (BLOCKING — wait for validation or edits) 7. **On confirmation:** set `meta.status: "confirmed"` 8. **Update context.md** 9. **Check consistency** with the deliverables that already exist Note: `data_model.drawio` and `data_model.html` are **exports**, generated on demand by the backend from `d