pipeline

# /pipeline — The full run Runs the entire estimation flow in a single session. For individual deliverables, use the modular skills. **UX: chat stays conversational.** No deliverable is ever presented as a markdown table in chat — every phase writes the JSON FIRST (`meta.status: "draft"`), the user reads it in the viewer, and then it gets discussed in words, citing elements by id. On confirmation `meta.status` becomes `"confirmed"`. 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. **The skills this composes:** - `/meeting` — merges a meeting transcript into the brief - `/data-model` — data model (tables, relations, integrations) - `/estimate` — the work plan in days - `/dev-tasks` — the granular dev task breakdown - `/mockup` — component-based mockup Each phase below follows the protocol of the corresponding skill. Read that skill's SKILL.md when you need the detail — this file is the running order, not a replacement for them. --- ## THE FLOW ``` [MEETING, if needed] → INTAKE → Q&A [STOP] → DATA MODEL [STOP] → ESTIMATE [STOP] → DEV TASKS → MOCKUP ``` Three blocking points: Q&A, data model, estimate. The user confirms before you go on. --- ## PHASE 0 — MEETING (conditional) Read `projects/{project}/.project.json`. If `source` is `"discovery"` and `projects/{project}/meetings/` holds transcripts not yet merged into `brief.json` (compare with `changelog[].source`), the pip...