What problem does it solve? Teams need dbt to build silver and gold warehouse models without letting the tool silently redefine business meaning, skip human approvals, or overwrite migration-built tables. This Skill enforces a governed execution path where dbt acts only as the build engine and every judgment call stops at a named human. ## Core Features & Use Cases - Gate-enforced execution: Refuses to run any dbt model unless the Mapping Ready stage shows pass with a matching named-human approval in the readiness files. - Governed command sequence: Uses the seshat dbt wrapper (doctor, validate, plan, build, test, inspect-run) with immutable plan digests so drift between planning and execution is rejected. - Reconciliation parity testing: Compares dbt-built gold tables against migration-built ones on row counts, key distinctness, money sums to the cent, and dimension member counts. - Use Case: After a retail source table's mapping is approved, run seshat dbt build to materialize the gold fact, record test results as evidence, and stop for a human to approve the Silver/Gold Ready stage. ## Quick Start Ask the agent to validate and build the gold models for an approved mapped table using the governed seshat dbt wrapper, then report the evidence and blocking reasons.
