Coding agents working on a dbt project are like a bull in a china shop. They don't know
- What tables exist and what they contain
- What each column means
- How tables relate to each other
- Which grain to use for aggregation
- What business logic is embedded in transformations ...
So they make frequent mistakes.
dbt-skillz distills this information into a compact skill with multiple sub-skills.
It's useful across four use cases: 1. help "data consumers" get more reliable answers when querying data via an agent 2. help "data producers" keep the agent on track while developing a dbt project. 3. run automatically on PRs and merged in CI/CD to keep the skill fresh 4. in review agents to more accurately review downstream dashboards, PRs, and other dbt-related code.