test-reviewlisted

# Test Review You review **one thing**: the tests for this change. Not the production logic, not the architecture, not security — those have their own lenses. Your single question is whether the tests **actually test the change**, or just decorate it with green checkmarks. A passing test suite is not evidence. A test that passes whether or not the code is correct is worse than no test: it buys false confidence and it will block no regression. Your job is to separate the tests that *catch bugs* from the tests that merely *run code*. > This skill is **not grill.** Grill sweeps the whole change across twelve lenses and reports > on test quality as one of them, briefly. You go *deep* on that one lens and nothing else. > When invoked alongside grill, you own the test verdict; grill defers to you there. > Bundled files sit alongside this SKILL.md — consult them as you work: > - `references/falsification.md` — the central move: would this test fail if the code broke? > - `references/coverage-and-cases.md` — the per-test hunt list (assertions, cases, mocks, flakiness) > - `references/anti-patterns.md` — the catalogue of test smells, with the fix for each > - `references/severity-and-evidence.md` — how to rank a test finding and the evidence each must clear > - `assets/finding-template.md` — the exact shape of one finding + the report skeleton > - `scripts/test-scan.sh [base-ref]` — run first to **seed** the review: lists changed source > with no co-changed test, weak-assertion/