Video 06 — Pat breaks the rewrite into 10 stories
Protagonist: Pat (Product Engineer) Arc: Setup Target length: 2:30 Apps in frame: Issue Manager · Context Manager · Workflow Studio
Story beat
Pat knows Buildbarn and knows what buildbarn-forms needs to do. The hard part isn’t writing the epic — it’s turning one sentence (“finish the React library”) into ten acceptance-criteria-complete stories an AI agent can execute against. RobOS does the translation: Pat writes the epic, connects the right context sources, and the AI drafts the ten stories with real technical scope.
Pre-seeded state
- Inherits the settings.json Dana produced in Video 01.
- Fixture F4 not yet created — this video creates it.
buildbarn-formsknowledge graph (~/source/hermetiq/hermetiq-genai-agent) is available locally; Pat will add it as a Context Manager source.
Scene list
- Cold open — Pat logs in. Dev Central shows “no tasks assigned”. Context Manager is empty.
- Context wiring (0:20–0:50) — Open Context Manager. Add three sources: the
hermetiq-genai-agentknowledge graph (local folder), thebuildbarn-formsrepo, and thebuildbarn-forms-protorepo. Scope all three to Issue Manager. - Epic creation (0:50–1:30) — Open Issue Manager. Click New Issue → type
bbf→ Workflow Studio suggests type “Feature” or “Epic” based on size → select Epic. Paste in Pat’s epic description (pre-drafted fixture). Tag withscope:buildbarn-forms. - AI breakdown (1:30–2:15) — Click AI Breakdown button on the epic. Panel opens. Pat types: “Break this epic into implementation-ready stories based on the knowledge-graph refactor plan and the current state of the buildbarn-forms repo.” Click Generate. Claude streams a list of ten stories, each with: title, description, acceptance criteria, estimated size, dependencies. Pat reviews, tweaks one title, clicks Save all as child stories.
- Close (2:15–2:30) — Task Board opens automatically showing the 10 new stories in the backlog column. Narration lands: “Ten stories, ten engineering videos. Let’s build.”
Narration cues (draft)
| # | At | Text |
|---|---|---|
| 1 | 0:00 | Pat’s the product engineer — knows the problem, knows Buildbarn, knows what the library needs to do. |
| 2 | 0:20 | Before writing the epic, Pat gives the AI the right context. Three sources: the knowledge graph, the draft repo, and the proto types it depends on. |
| 3 | 0:55 | Now the epic itself — one paragraph about what “done” looks like for buildbarn-forms. |
| 4 | 1:35 | AI Breakdown takes the epic, reads the context, and drafts implementation-ready stories with acceptance criteria. |
| 5 | 1:55 | Ten stories, each scoped so a developer can pick one up and know exactly when they’re done. |
| 6 | 2:20 | Every one of those stories becomes a video in this series. Let’s build. |
Blockers / ready-checklist
- The “AI Breakdown” button in Issue Manager — does it exist as of v0.0.5, or is it one of the features we’re narrating past? Verify. If not built, decide to build it or use a canned response flow similar to our workflow-studio gh-copilot fixture.
- Pre-draft the epic description (prose the viewer sees Pat paste in) and the ten stories the AI should produce. Save as
packages/robos-test/sandbox/data/model-problem/bbf-epic.md+bbf-stories.json. - Context Manager’s per-source scoping to Issue Manager must be real — verify the scope picker in the app.
Deliverables produced by this video
- Fixture F4 — epic + 10 stories populated in the BBF Jira project and in
~/.config/robos/settings.json. - Backlog is now ready for video 04 (Alex’s onboarding) and 05 (first engineering pick-up).