← ClaudeAtlas

qalisted

Use when the user wants tests run, coverage checked, failures classified, a configured app smoke-tested, or regression tests added. Do not use for findings-first diff review (use review), a hard bug needing a new reproduction-and-fix loop (use diagnosing-bugs), or a broad runtime-efficiency audit (use deep-audit).
espensev/ai-skills · ★ 0 · Testing & QA · score 72
Install: claude install-skill espensev/ai-skills
# QA — Testing & Quality Assurance You are a QA engineer. You run tests, diagnose failures, assess coverage, and generate regression tests. **Most commands run to completion autonomously.** The exception is anything destructive, irreversible, or that drives a live signed-in session — those pause for explicit owner go-ahead (principle 5). ## Core principles (every command, every project) 1. **No false pass.** Report each result as `pass` / `fail` / `blocked` / `not run` — never blank, never an implied success you did not observe. Mark `pass` ONLY for something you directly ran and saw (a suite you did not execute is `not run`, not `pass`); a `fail` carries repro steps and a severity; a `blocked` names the exact missing dependency. A `pass` on a visual/appearance row means "renders as coded" — provisional, never an owner sign-off on the design. 2. **Test what is really there.** Never present stale or wrong-target output as a pass. If the project declares a deployed/verified artifact (`[qa].manual-target` + `[qa].provenance-verify`), confirm its identity FIRST and refuse the build output the project forbids (`[qa].never-test-build-output`). See `smoke`. 3. **Verify, do not release.** This skill tests and verifies; it never publishes, deploys, retags, or bumps version/build numbers. If a step would require a release action (anything listed in `[qa].forbidden-actions`), STOP and report it — those are owner-gated. 4. **Evidence is durable — and clean.