These tasks use package macros to build output, then add guarantees around that output. Pin the dependency, inspect the SQL it generates, and keep the distinction between an identifier, a display field, and a data test.
A surrogate key needs an intentional NULL policy; a pivot needs an explicit measure.
The date scaffold crosses a month boundary, and the capacity rule includes a fully reserved route. Both cases test boundaries that ordinary examples can hide. The final build uses actual pre- and post-hooks to write an audit record after materialization.
Keep that side effect separate from the model SELECT, and verify the intended order from the project configuration.