Mutation Testing Advisor

---
name: mutation-testing-advisor
description: Manually introduce small, targeted logic mutations (flip a comparison operator, shift a boundary by one, invert a condition) into already-tested code and check whether the existing tests fail. Use to spot-check whether coverage is real assertion strength, not just line execution.
---

# Mutation Testing Advisor

Line coverage tells you code *ran* during a test. It doesn't tell you the
test would notice if that code were wrong. This skill checks the second
thing, cheaply, by hand.

## Process

1. Pick a function with test coverage that matters (business logic,
   validation, a calculation) -- not a trivial pass-through.
2. Introduce ONE small mutation at a time, in a scratch copy or a
   throwaway branch, never in the real diff:
   - Flip a comparison operator (`<` to `<=`, `==` to `!=`).
   - Off-by-one a boundary constant.
   - Invert a boolean condition.
   - Change a `return` value to a different valid-looking value.
3. Run the existing tests against the mutated code.
4. If all tests still pass, that's a **surviving mutant** -- the test
   suite has a gap at exactly that logic, regardless of what the coverage
   report says.
5. Revert the mutation. Repeat for a handful of the function's important
   branches -- this is a spot-check, not exhaustive coverage of every
   possible mutation.

## Output

For each mutation tried: what was mutated, whether a test caught it, and
for every surviving mutant, what test (or assertion inside an existing
test) would need to be added to catch it.