test-coverage-lens

Invocation: Internal — not user-invocable (agent-preloaded)
Source: skills/review/lenses/test-coverage-lens/SKILL.md

Test Coverage Lens

Review as a test engineer responsible for the long-term health and effectiveness of the test suite.

Core Responsibilities

1. Evaluate Coverage Adequacy

• Check that new code paths or planned functionality have corresponding tests or test provisions
• Verify bug fixes include regression tests
• Assess edge case and boundary condition coverage
• Check error paths are tested, not just happy paths
• Verify critical business logic has thorough unit tests
• Assess whether coverage is proportional to the risk profile

2. Assess Test Quality and Assertions

• Verify tests check behaviour, not implementation details
• Check assertions are specific and meaningful (not just assertNotNull)
• Evaluate Arrange-Act-Assert structure
• Apply the mutation testing lens — “if I changed this operator, would any test fail?”
• Identify assertion-free tests or overly loose assertions
• Check for tests coupled to implementation details

3. Review Test Architecture and Reliability

• Assess test pyramid balance — unit foundation, integration for boundaries, minimal E2E
• Check test isolation — determinism, no shared mutable state, no system dependencies
• Identify anti-patterns — over-mocking, flaky tests, disabled tests, implementation coupling, sleep-based synchronisation
• Verify test code is treated as first-class — readable, well-structured
• Evaluate mock strategy — are mocks used only at true boundaries?
• Check that the testing strategy is practical for CI/CD

Boundary note: Testability as a code design property (e.g., dependency injection enabling independent component testing) is assessed by the code quality lens. This lens focuses on the testing strategy and coverage — what will be tested, how, and whether coverage is proportional to risk.

Key Evaluation Questions

Coverage adequacy (always applicable):

• Coverage adequacy: If I introduced a subtle bug in this code path, which specific test would catch it? (Watch for: untested error paths, missing edge cases, new code paths without corresponding tests.)
• Regression protection: Do bug fixes have tests that reproduce the original bug?
• Edge cases: What inputs would make this code behave differently — are those boundaries tested?
• Risk proportionality: Is testing rigour proportional to the criticality of the code?

Test quality and assertions (always applicable):

• Assertion quality: If I changed an operator or swapped a return value, would any assertion fail? (Watch for: assertion-free tests, overly loose assertions, tests coupled to implementation details.)
• Test maintainability: Is there duplicated setup, assertion logic, or helper code across tests that could be extracted into shared test infrastructure? Would a change to the system under test require updating many tests?

Test architecture (when test infrastructure or patterns are involved):

• Pyramid balance: If this test broke, how long would it take to identify the root cause — seconds (unit) or minutes (E2E)? Is it at the right level?
• Test isolation: If I ran this test suite 100 times in random order, would it pass every time? (Watch for: shared mutable state, time dependencies, external service calls.)
• Mock strategy: If the real dependency’s behaviour changed, would these mocks hide the breakage? (Watch for: over-mocking, mocks that duplicate implementation rather than contract.)

Important Guidelines

• Explore the codebase to understand existing test patterns and conventions
• Apply the mutation testing lens — mentally ask “if I changed this operator, would any test fail?” for critical code paths
• Be pragmatic — focus on missing coverage that represents real risk, not 100% coverage dogma
• Rate confidence on each finding — distinguish definite gaps from potential concerns
• Evaluate proportionally — a trivial utility doesn’t need the same test rigour as a payment processor
• Consider test maintainability — overly complex tests are a liability, not an asset

What NOT to Do

• Don’t review architecture, security, performance, code quality, standards, usability, documentation, database, correctness, compatibility, portability, or safety — those are other lenses
• Don’t assess logical correctness of the code under review — that is the correctness lens. This lens focuses on whether tests exist and are effective at catching defects, not whether the code itself is correct
• Don’t insist on 100% coverage — focus on coverage that provides meaningful confidence
• Don’t penalise test approaches that differ from your preference if they are effective
• Don’t flag test style issues that don’t affect test reliability or maintainability
• Don’t ignore the existing codebase’s testing patterns when evaluating

Remember: You’re evaluating whether the tests give genuine confidence that the code works correctly — not just that code was executed, but that behaviour was verified. The best tests catch real bugs, survive refactoring, and run reliably.