Quick Start
This page takes you from zero to a working CAP Pro feature in five commands. It assumes you have CAP Pro installed (npx cap-pro@latest) and a project to work in.
1. Initialise the project
In your project directory, run:
/cap:initThis creates:
.cap/— ephemeral session state and project memoryFEATURE-MAP.md— single source of truth for features and acceptance criteria.cap/stack-docs/— auto-fetched library documentation via Context7
/cap:init runs brownfield analysis on existing codebases — it will scan your code for existing patterns and pre-populate the Feature Map.
2. Discover features
/cap:brainstormThis launches an interactive conversation with the cap-brainstormer agent. It:
- Asks targeted questions to understand what you want to build
- Clusters features into logical groups
- Surfaces dependencies between features
- Produces structured Feature Map entries with acceptance criteria (ACs)
Output: FEATURE-MAP.md populated with feature entries in state planned.
3. Build the prototype
/cap:prototypeThe cap-prototyper agent:
- Reads the active feature from
FEATURE-MAP.md - Confirms the ACs with you (one round of "is this right?")
- Builds working code, embedding
@cap-featureand@cap-todotags - Marks the feature
prototypedin the Feature Map
The tags are how CAP Pro keeps the plan in sync with the code. Example:
// @cap-feature(feature:F-Hub-Spotlight, ac:1) Display spotlight banner with rotating slides
// @cap-todo(ac:F-Hub-Spotlight/AC-3, risk:medium) Hover-to-pause needs reduced-motion fallback
function SpotlightBanner({ slides }: Props) {
// ...
}4. Test it
/cap:testThe cap-validator agent (in MODE: TEST) writes runnable tests against the Feature Map ACs using RED-GREEN discipline:
- Write the failing test first (RED)
- Run it — confirm it fails for the right reason
- Make it pass (GREEN)
The test framework is auto-detected (vitest, node:test, jest, pytest, …). Once all ACs are green, the feature transitions to tested.
5. Review & ship
/cap:reviewcap-validator (in MODE: REVIEW) does a two-stage review:
- Stage 1 — AC compliance: every AC in the Feature Map has a green test? Every
@cap-todois closed? - Stage 2 — Code quality (only if Stage 1 passes): security, performance, maintainability, anti-patterns.
Once both stages pass, the feature moves to shipped and you commit + tag.
What's next?
- Stuck? →
/cap:debugruns the scientific-method debugger with persistent state across context resets. - Lost track? →
/cap:statusshows the feature dashboard (or/cap:startre-loads your last session). - Different team member taking over? → see Multi-User Workflow.
- UI/design sprint with lots of small tweaks? → see Frontend Sprint Pattern.
The auto-trigger contract
You don't have to type the slash commands. CAP Pro recognises the workflow moment and auto-invokes the right agent:
| Situation | Auto-action |
|---|---|
| You describe a new feature without a Feature Map entry yet ("we need X") | /cap:brainstorm |
Feature in state planned, you say "let's build it" | /cap:prototype |
Feature in state prototyped with open @cap-todo tags, you say "iterate" | /cap:iterate |
Feature in state prototyped but not tested yet | /cap:test |
Feature in state tested, you say "ready to ship" | /cap:review |
| You report a bug or "this works locally but not in prod" | /cap:debug |
| You ask "where are we?" / "show me the dashboard" | /cap:status |
The slash commands exist as explicit power-user triggers. In normal conversational use, just describe what you want and CAP Pro picks the right tool.