Mutation testing

Skill jagreehal/jagreehal-claude-skills/skills/mutation-testing

Uses mutation testing to measure test effectiveness and close gaps that coverage alone misses. Use this skill when evaluating whether tests would catch real faults, tuning Stryker, or improving weak suites after green coverage. Do not use when/for writing the first tests (use writing-tests) or load/chaos testing (use performance-testing).From its SKILL.md

Install

npx -y skills add jagreehal/jagreehal-claude-skills --skill mutation-testing

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

3 stars3 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

3.7 KB, as published. Nobody here has run it

Mutation Testing

Critical rules

Coverage = execution. Mutation = detection. Prefer Stryker over manual mutation for JS/TS.
Run at the PR-readiness / end-of-phase gate — never inside each RED-GREEN increment.
Diff-scoped runs for branches; full runs for setup, CI, or baseline work.
Mutate production source only — never tests, fixtures, snapshots, generated, .d.ts, build output.
Every survivor: kill with a behavior test, classify equivalent, or ask the user.
New tests must fail against the mutant for the right reason before restoring production code.
Assert exact observable behavior — not toBeDefined / weak inequalities.
Before operators, survivors, or WRONG/CORRECT examples, read the matching resource below.

Workflow

Inspect — Find package manager, runner, Stryker config; git diff <base>...HEAD --name-only. Monorepo: smallest affected package first.
Setup if missing — npm init stryker@latest; Vitest plugin; coverageAnalysis: perTest; mutate src/**/*.ts excluding tests. Add scripts: mutation, mutation:incremental, mutation:diff. Before editing config details, read references/setup.md.
Run — Prefer mutation:diff for branch feedback. Full mutation for first setup / shared infra / CI gates.
Triage — Killed: ok. Survived: strengthen the assertion or change the input (equal original/mutant results for the current input means no assertion can kill it), unless equivalent. No Coverage: add behavior test. High-value code (money, permissions, safety): fix immediately. Subtle/unspecified: ask user with concrete choices.
Kill survivors (TDD) — Keep mutant → write smallest failing behavior test → restore code → verify pass → re-run scoped mutate then diff. Assert returns, persisted state, events — not implementation details.
CI — Start report-only; add failing thresholds after a baseline. Persist HTML/clear-text artifacts. Incremental locally; force full periodically.

During RED (cheap): use mutator heuristics mentally — see references/mutator-rules.md. Manual fallback only when Stryker can't target the code.

Resources

references/mutator-rules.md — operators, heuristics, equivalent mutants, manual procedure. Read when planning tests or triaging survivors.
references/setup.md — Stryker config, scripts, diff scoping. Read when adding or tuning the harness.
references/examples.md — weak coverage vs boundary-killing tests. Read when strengthening survivors.

Validation

Harness is Stryker (or documented manual fallback)
Run timed at PR gate, not per TDD increment
Diff-scoped for the branch; mutate production only
Survivors killed, marked equivalent, or escalated to user
Killer tests failed against the mutant first; assert exact behavior

Constraints

Not for authoring first tests (writing-tests) or load/chaos (performance-testing).
Adjacent: tdd-workflow (cycles before this gate), testing-strategy (pyramid placement), verification-before-completion (include mutation report in evidence).