Choose how to add tests to a pure module

Your team wants about 80% test coverage on a stable, well-typed Python module of pure functions: no I/O and no shared state. One engineer sets up an autonomous agent to "add tests until coverage hits 80%", running pytest in a loop. Another wants to generate the whole test file in one prompt from the module's source.

Which approach fits this module, and what is the real risk you are managing?