DYLA-ai: Skill for Claude Code

/meeting — From transcript to brief

Discovery is the phase where you interview the people who will use the thing: how they work today (usually spreadsheets, email and a couple of macros) and how that process should look once it is a real application. The brief is what comes out of those conversations.

This skill does one job: turn a transcript into a document. It does not rewrite the brief from scratch at every meeting — it grows it.

---

UX: chat stays conversational. Never paste the brief, its chapters or lists of questions into chat: the user reads them in the viewer. In chat you say what changed, citing chapters by id and title (e.g. "C3, Permissions") and questions by id (e.g. "Q7"). 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. Check the project source — this skill only applies to source discovery
2. Find the transcripts not merged yet (meetings/ vs the changelog of brief.json)
3. Read the transcript, the current brief, context.md, the existing questions and people
4. Merge into the brief incrementally (meta.status: "draft", version +1)
5. Extract open questions, decisions, people, contradictions
6. Say in words what changed (BLOCKING — wait for confirmation or edits)
7. On confirmation: meta.status: "confirmed"
8. Update context.md

---

PHASE 1 — CHECK DEPENDENCIES

Read projects/{project}/.project.json. If source is not "discovery", say so and ask in a ```questions block (format in .Codex/prompts/questions_format.md):

```questions
[
  {
    "id": 1,
    "q": "This project starts from a brief that was handed to us: it is an input document and this skill does not rewrite it. Do you still want this meeting's decisions and open questions kept on record, in context.md and questions.json?",
    "options": ["Yes, keep them in context.md and questions.json", "No, nothing else needed"]
  }
]
```