When to Apply This Skill
-
A developer submits an existing test file for quality audit
-
A team lead wants to assess test quality before a release or merge
-
A CI/CD gate requires a quality verdict before a PR is approved
-
A test suite generated externally needs validation before use
Atomic Skills to Load First
Read these files before executing any step:
../unit-test-analyzing-code-coverage/SKILL.md
— Coverage type definitions, severity classification (CRITICAL / HIGH / MEDIUM / LOW), and gap report format — the reference standard for coverage dimension review
../unit-test-generating-unit-tests/SKILL.md
— AAA pattern, naming conventions, correct mocking structure, and exception testing patterns — the reference standard for judging test quality under review
Execution Steps
Read references/index.md before executing any step.
Step 1 — Receive Input
Accepted inputs (one or more):
-
Test file(s) to review (required)
-
Source code the tests are written against (recommended)
-
Coverage report: JSON (coverage.py / Istanbul), XML (JaCoCo / Cobertura)
-
Requirements or user stories the tests should satisfy
Step 2 — Invoke test-unit-review-agent
Pass all inputs to ../../agents/unit-testing/test-unit-review-agent.md .
The agent evaluates across 6 dimensions:
-
Coverage Completeness — all paths, branches, and edge cases addressed
-
Test Correctness — assertions verify the right things, no false positives
-
Test Independence — no shared mutable state, no order dependency
-
Naming and Readability — self-documenting names, AAA structure visible
-
Framework Compliance — correct imports, decorators, no deprecated APIs
-
Security and Safety — no real credentials, PII, or live network calls
Step 3 — Deliver Review Report
Always emit a structured report containing:
-
Coverage Score (X/10) and Quality Score (X/10)
-
Passed checks list
-
Issues table: Severity | Location | Issue | Suggested Fix
-
Recommended additions (specific missing test cases by name)
-
Final verdict
Output Deliverables
review_report.md <- full quality report with verdict revision_instructions.md <- (REQUIRES_REVISION only) annotated fix list ready to pass directly to test-unit-gen-agent