project-context

Invoke the project-context skill in data mode targeting your repository path. It reads the database schema, API routes, backend and frontend code, and existing docs, then writes .context/business/business-data-map.md with entities, flows, state machines, and integrations.

Run the skill in test-plan mode after generating the business data map, which is a hard requirement. It scores flows by business impact, breakage likelihood, and blast radius, then writes .context/master-test-plan.md with a ranked risk map and pre-release checklist.

Data mode maps entities, flows, and state machines; features mode builds a feature catalog with CRUD matrix and endpoint inventory; api mode narrates the business-level API story including auth tiers and critical journeys. Each mode writes a separate artifact and loads only its own reference.

No. When an artifact already exists, the skill enters UPDATE mode: it generates a candidate version, shows a diff summary, and waits for explicit approval before overwriting. Missing artifacts are created directly after analysis completes.

Test-plan mode stops with a hard dependency error and instructs you to run data mode first, since risk reasoning requires the data map. A missing feature map is only a soft dependency — the run continues and logs the limitation under Discovery Gaps.

Use refresh-all only when you explicitly want all four artifacts regenerated. It runs data, features, api, then test-plan sequentially, closing each mode's validation and approval gate before loading the next, and stops on any hard failure or rejected overwrite.