← Journey PHASE 5b — Design system: client design system + Penpot bidirectional

CLI Loop Documented — Brief to Penpot Frame in One Pass

Photo by NIKHIL on Unsplash

The full CLI workflow is now written up — every command from briefv2 to penpot-token-sync, with a walkthrough that covers the complete loop from client brief to refined Penpot frame.

The documentation section now has a user guideline covering all six CLI commands in Phase 5b. This is the reference I’ll hand to anyone running the design loop without wanting to reverse-engineer the workflow from the scripts themselves.

What shipped: a single doc that sequences briefv2, penpot-pattern, penpot-extract, penpot-sync, penpot-review, and penpot-token-sync — when to call each one, what it expects, and what breaks if you skip an earlier step.

dani-brain/wiki/design
├─ cli-loop.md new Full loop walkthrough: brief intake to refined Penpot frame
├─ commands/briefv2.md new Input schema, flags, output contract
├─ commands/penpot-pattern.md new Pattern selection logic and override flags
├─ commands/penpot-extract.md new Frame extraction, asset paths
├─ commands/penpot-sync.md new Sync direction, conflict rules
├─ commands/penpot-review.md new Review gate — what must pass before token sync
└─ commands/penpot-token-sync.md new Token push to repo, diff output

The walkthrough in cli-loop.md runs a full example: client brief in, pattern selected, frame generated, extracted, reviewed, tokens synced to repo. Each step links to its command doc. No inline tutorial — just enough context to know what the command is doing and why it’s in that position.

The Figma Dev seat option stays deferred. The REST API covers what’s needed right now and penpot-extract doesn’t have a gap that justifies $25–45/mo at this stage.

What this unblocks: With the loop documented, the next two tasks — data ingestion sub-workflow and template iteration — can be built against a stable command surface without the loop changing under them.