mutation-sweeplisted

<!-- mutation-sweep-version: 3 --> # Mutation sweep **A survivor list is a triage queue, not a defect list.** That is the posture everything below serves. Line coverage certifies that lines ran; a mutation sweep certifies that the tests would notice if those lines were wrong. The motivating evidence (issue #67): in one audited session every code-level HIGH sat on an already-covered line, and a nine-second sweep of one module found the two real defects four review rounds had missed. This skill deliberately overlaps the code-reviewer agent's "Assertions that cannot fail" dimension, which is the CANONICAL statement of the five rotten-green shapes AND of the division of labour between the sweep and the reviewer. Run both; neither replaces the other; consult that dimension rather than any restatement. ## Tool selection (criteria, not winners) Adopt the STACK'S maintained engine; never build one (an engine is exactly the executable this kit's own discipline would then demand contract tests for). Choose by criteria, which forge-adapt maps onto the project's stack at install time: - **Incremental runs and test-targeting** (only re-test mutants whose covering tests changed): the difference between a sweep people run and one they abandon. - **Operator breadth and build-tool integration** where the suite is framework-heavy. - Ecosystem exemplars the criteria map onto: `mutmut` / `cosmic-ray` (Python), `Stryker` (JS/TS/C#), `PIT` (JVM), `cargo-mutants` (Rust), `go-mutesting`/`g