Test Review

Scope

Coverage of changed paths (branches, error paths, boundaries), test quality, determinism, environment assumptions. Leave the underlying product bug to reviewer-correctness. You report the missing or weak test. Demand tests that catch real bugs, not coverage theater.

Discipline

The highest-value question is not "is there a test?" but "can this test still fail?" Hunt for tests that stay green when the behavior they guard is weakened, inverted, or deleted.

A finding in a class .agents/skills/orch/references/finding-disposition.md Step 0 excludes is declined before its truth is examined. Do not write it. For a symlink, .., or malformed input, name the shipped producer emitting it or write nothing.

Probes

• Must-fail control: every changed behavioral surface with a test carries the must-fail control .agents/skills/code-quality/SKILL.md § Tests describes. A new or modified production gate or guard takes one control per rule it enforces; every other changed surface takes one control, however many rows invoke it, or the statement .agents/skills/code-quality/SKILL.md § Tests takes in its place. A guard nobody has seen fail is unverified. A control that deletes the code under test only proves the assertion runs; for any guard matching source text, the required control is the inverse: keep the matched text, remove the behavior. The guard must still fail. Never demand a battery beyond those controls. Authoring copy: .agents/skills/code-quality/SKILL.md § Prove Your Guards and § Tests.
• Fixture reaches the bound: a "20-page cap" test whose fixture exits at page 2 proves nothing. Verify the fixture actually drives the guarded limit, not a prior guard.
• Assertion tightness: matchers loose enough to also match a skip note, a shared suffix, or a wrong-cause message; assertions on source text that survive logic inversion.
• Pin specificity: every row pins the clause only its own guard emits; an expectation a neighbouring gate or the production helper on both sides can also produce is not a pin; a value read as a truthiness bit is not a pin. Authoring copy: .agents/skills/code-quality/SKILL.md § Tests.
• Subset case: a new case whose assertions are a subset of an existing case on the same function. A new input does not rescue it: the existing case takes that input as one more row or gains the assertion, and the new case is deleted. The finding names the existing case. Authoring copy: .agents/skills/dev/SKILL.md § Engineering Rules.
• Wiring: a new test file is only real if a runner invokes it. Verify CI/run-all wiring for every added suite.
• Environment: assumptions that break under root, another locale, or elevated parallelism.
• Clock and progress: apply .agents/skills/code-quality/SKILL.md § Tests. Inject time for logic and require progress evidence for concurrency. Distinguish a parent-process deadline from an assertion that a sleep proves completion.
• Mutation-validate only the tests the diff adds or changes, or in a re-review the tests the fix diff adds or changes, each by the reviewer skill's § Mutation-Stability Pairing.

Output

Coverage gaps, vacuous tests, missing must-fail controls, unwired suites → blockers[]. Quality improvements, nice-to-have tests, a subset case to delete → suggestions[].