Use the verification cluster when example-based tests cover the obvious cases but the bug class is broader than the corpus.
- Hypothesis — when a property should hold across an input space
(round-trip, oracle, invariant). Mirrors the rust-skill
proptestdeep dive. - CrossHair — when symbolic execution can prove or disprove an
assertion over a small pure function. Best on parsers, codecs,
finance code, and refactors validated with
diffbehavior. - mutmut — when the suite passes consistently and the question is whether the tests would notice a regression.
Run Hypothesis on every CI run; run CrossHair and mutmut on slower cadences.