# Review Test Quality

> Test quality review for meaningful assertions, isolation, determinism, negative cases, and unit, integration, contract, and end-to-end balance. Use for test-strategy, flaky-test, or coverage-confidence questions.

- Skill: `kenlck/review-test-quality` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add kenlck/review-test-quality`
- Raw SKILL.md: https://api.skillmd.com/api/skills/kenlck/review-test-quality/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Security
- Author: kenlck (https://skillmd.com/u/kenlck)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/kenlck/review-test-quality

---


# Review test quality

Assess one repository, a repository group, or an existing fleet evidence bundle. Judge the confidence tests provide, not the volume of test files or a coverage percentage alone.

## Run the lens

1. Read the [standalone lens protocol](../audit-software-fleet/references/standalone-lens.md) completely and prepare the supplied scope with lens ID `test-quality`.
2. Read [the test quality checks](references/checks.md) completely. Apply every check family relevant to the repository role, change risk, interfaces, and available execution evidence.
3. Compare tested behaviors and failure cases with production boundaries. Classify supported outcomes as strengths or findings; record unavailable execution, coverage, or flake evidence as unknowns and incomplete coverage.
4. Follow the protocol to account for every contract-registered check family, validate the result, ingest it, and hand back the canonical artifact path.

For raw collection or test execution, read and follow the [safety model](../audit-software-fleet/references/safety-model.md). Run tests only through explicitly authorized deep-mode adapters.

Read [scanner adapters](../audit-software-fleet/references/scanner-adapters.md) before interpreting test, coverage, or execution-tool output.

## Boundary

Own the confidence and defect-detection value of tests. Route pipeline enforcement to delivery and supply chain and runtime failure semantics to software reliability.

## Completion

**Complete when:** every in-scope repository has an exact commit or explicit unavailable state, every applicable check family is accounted for, every claim has bounded sanitized evidence at an appropriate confidence, gaps remain visible in unknowns and coverage, and the shared result validator succeeds.