ai-generated-ut-code-review_skill

Overview

This skill reviews AI-generated unit tests and scores their quality on a 0–10 scale. It evaluates coverage, assertion effectiveness, negative/edge-case handling, determinism, and maintainability. The output includes a numeric score, a risk level, and a prioritized must-fix checklist tied to concrete evidence.

How this skill works

It confirms tests exercise the intended business code and core paths, checks overall line coverage (requirement: >= 80%), and inspects assertions for behavior relevance. It verifies negative and edge cases, looks for flakiness (time, network, env), and assesses test structure and naming. Finally, it produces a score by dimension, assigns a risk level, and lists must-fix items with supporting evidence and quick remediation notes.

When to use it

• When evaluating AI-generated unit tests for release readiness
• When you need a numeric quality score and an actionable risk level
• When coverage looks high but assertion validity is unclear
• When you need a prioritized must-fix checklist for test debt
• When assessing whether tests are deterministic and isolated

Best practices

• Require overall line coverage >= 80% before marking coverage as acceptable
• Ensure each test includes at least one behavior-anchored assertion (not just non-null or log checks)
• Include negative, boundary, and exception-path tests for core logic
• Avoid external dependencies; mock or stub external systems and freeze time where needed
• Name tests clearly and group by behavior to improve maintainability

Example use cases

• Score a batch of AI-generated tests to decide if they can be merged
• Identify missing negative cases when coverage is reported as high
• Produce a must-fix checklist for engineers after an automated test generation run
• Assess whether generated tests are flaky due to time, network, or environment reliance
• Compare multiple AI test outputs and pick the safest candidate for production