Skills

The plugin ships three skills. Claude picks them up automatically; builders never invoke them by name.

sprig-publish

Triggers on: “share this solution”, “publish this app/dashboard”, “deploy to sprig”. Does: validates the app has Entra auth configured (refuses otherwise), ensures machine configuration, authenticates the builder (device code, first time), calls the platform to provision/update, uploads content with a scoped token, reports the URL and the owner-only status. Re-publishes update in place.

sprig-fabric-data

Triggers on: “build a dashboard with data from …”, Fabric/lakehouse data requests. Does: schema introspection against your Fabric GraphQL endpoint (as the builder), query validation with real calls, then generates viewer-identity frontend code — including correct pagination and the hard rule that data is never baked into the artifact. Declares semantic models out of scope (roadmap).

sprig-ops

Triggers on: operator asks — “pending share requests”, “approve that request”. Does: lists the queue and submits approve/reject decisions (requires the Ops app role).

Under the hood

Skills drive small command-line tools shipped with the plugin (lib/publish.mjs, lib/ops.mjs, lib/fabric-query.mjs, lib/init.mjs). They are equally usable directly — every skill action has a plain CLI equivalent, documented in the pages above.