mutation-testing

Invoke the skill with a file, directory, or description of scope, such as a specific module. It applies 3-8 deliberate mutations per file, runs your test suite after each one, reverts via git checkout, and reports which mutations survived as test gaps.

Mutation testing introduces small deliberate bugs into source code and checks whether tests fail, measuring whether tests verify behavior. Code coverage only shows which lines executed, so fully covered code can still have tests that assert nothing meaningful.

Yes, the skill detects the test runner automatically by looking for pytest.ini, pyproject.toml, package.json test scripts, go.mod, Cargo.toml, Makefiles, or CI configs. If it cannot determine the test command, it asks the user directly.

Yes, the skill requires a clean git working tree before starting and reverts every mutation immediately with git checkout after each test run. It verifies the revert with git diff and never mutates test files, imports, or docstrings.

Survived mutations are reported with an explanation of the untested behavior and written to TODOS.md as critical [TEST] items. They are meant to be closed with the /tdd skill's red-green-refactor loop rather than ad-hoc test writing.

Avoid it when the test suite is already failing, when there are uncommitted changes to files in scope, or on boilerplate and trivial accessors with no meaningful logic. Slow test suites may also require running relevant subsets instead of the full suite per mutation.