diagramlisted
Install: claude install-skill Federico-Anastasi/DYLA-ai
# /diagram — Technical diagrams
Reads the project material and produces the technical diagrams: architecture views
(systems and integrations, SAI-style), operational workflows, dataflows (who touches
which data — also the base for data-privacy mapping), sequence diagrams for the
exchanges that need step-by-step precision.
**UX: chat stays conversational.** The deliverable is never drawn in chat — the user
sees and edits the diagrams on the canvas in the viewer, from `diagram.json`. In chat
you discuss in words, citing diagrams and nodes by id (e.g. the `crm` node in
`arch-overview`). When you cite a chapter of the brief, write `[[brief:Chapter title]]`.
---
## PROTOCOL
1. **Read context.md** — where is the project? (Diagrams already discussed? Decisions taken?)
2. **Read the material** — brief, data model, estimate, meetings (see PHASE 1)
3. **If `diagram.json` already exists:** ask "Edit it, or regenerate from scratch?"
4. **Q&A** (BLOCKING — 3 to 5 questions, no more): which diagrams, which scope
5. **Write diagram.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**
Note: `diagram.html` is an **export**, generated on demand by the backend from
`diagram.json` (`GET /api/projects/{project}/export/diagram.html`) — a single
self-contained file the client opens with a double click. This skill does not produce it.
--