What problem does it solve? AI-generated code often ships with example-based tests written by the same agent that wrote the implementation, propagating identical blind spots into both files. This Skill replaces that self-deception loop with verification against mathematical properties, invariants, and randomized input spaces. ## Core Features & Use Cases - Property-Based Testing (PBT): Extract algebraic invariants (round-trip, commutativity, idempotency, monotonicity) from specifications and validate them with randomized generators using tools like hypothesis or proptest. - Metamorphic Testing: Solve the oracle problem by asserting metamorphic relations (permutation, scaling, monotonicity, invariance) across perturbed inputs when expected outputs are unknown or expensive to compute. - Fuzzing & Hierarchical Verification: Build fuzzing harnesses for untrusted input boundaries and structure test suites in tiers from compiler gates through E2E and differential assertions. - Use Case: When implementing a serialization module, instead of hardcoding input-output examples, write a round-trip property test asserting decode(encode(x)) == x across randomized inputs, then fuzz the parser boundary for crash resistance. ## Quick Start Ask the agent to design a property-based test suite with metamorphic relations and a fuzzing harness for the module you are implementing.
