Audit, harden, or generate unit tests with a focus on mutation-resistance and durable quality — not just coverage %. Use when the user asks to improve or harden test quality, make tests less fragile/brittle, review tests for anti-patterns (error-message-substring asserts, private-symbol access, tautological constructor readbacks, recomputed-crypto expectations, over-mocking, missing boundary tests, unrolled cases that should be parametrized), generate a high-quality suite for an untested module, or generally "make my tests better". Measures a suite against a validated quality scorecard and iterates until it plateaus, holding coverage as a non-regression floor and REPL-verifying library assumptions. Works across stacks — Python/pytest (validated), JavaScript/TypeScript (Jest, Vitest, Mocha/Chai, node:test), Go, Kotlin (kotlin.test/JUnit5/Kotest), and Swift (XCTest/Swift Testing/Quick) — with a per-language scorer; the rubric is the same everywhere.
Turn a test-quality scorecard JSON into a self-contained interactive HTML readout — hero stats, a per-axis Win/Tie/Loss distribution chart, and the full per-suite matrix. Use when the user wants to visualize benchmark/scorecard results, regenerate the results dashboards in docs/, or render a {baselines, arms} scorecard (as emitted by the benchmark's score_quality.py / score_cross_language.py, or the test-quality skill's scorer) as a shareable web page. Produces one standalone .html file (Chart.js via CDN, no build step).