orchestra-blueprints

Build a new dlt source that lands API data into MotherDuck, then ship it through Orchestra. Use this when the user wants to "pull X from <API> into MotherDuck" or "add a new dlt source and run it on Orchestra". Walks the full lifecycle on a feature branch — author the source, register an Orchestra pipeline that invokes it, trigger the pipeline against that branch, and poll until done.

Publish or update a MotherDuck Dive by inlining a local .tsx file into a `MD_CREATE_DIVE` / `MD_UPDATE_DIVE` SQL call sent through the MotherDuck MCP (`mcp__MotherDuck__query_rw`). Use this when the user says "publish the dive", "ship this dive to MotherDuck", or "update dive <id>". Reads the TSX from `.dive-preview/src/dive.tsx` (override-able), SQL-escapes it, executes via MCP, and records the returned dive UUID in `.dive-preview/dive_id.txt`.

Reads Snowflake Metadata to construct a pipeline with DQ Rules

Given a failed Orchestra pipeline run ID, this skill automates the full workflow of diagnosing the failure, applying a code fix on a new GitHub branch, and re-running the pipeline on that branch.