grade-tests

Provide the Skill with the list of changed test methods, file paths, or the PR diff, and it grades each test on assertion strength, structure, and anti-pattern hygiene. The output is a markdown table with letter grades, score bands, and one-line notes ready to post as a PR comment.

Grading supports MSTest, xUnit, NUnit, and TUnit for .NET, plus pytest, Jest/Vitest, Go testing, JUnit-style Java, Ruby, Rust, Swift, Kotlin, PowerShell, and C++ frameworks. Language-specific assertion conventions are loaded from extension files before scoring.

No, the Skill rejects ambiguous requests without an explicit scope and asks for a method list, file paths, or a diff. For full-suite audits, use a test-quality-auditor agent or a test-anti-patterns skill instead.

Each test starts at A and receives deductions only for observable issues across three sub-grades: assertion strength (45%), anti-pattern hygiene (30%), and structure (25%). The weighted average maps to a letter grade, capped at the worst sub-grade, and is reported as a score band rather than a precise number.

Grades drop for tests with no assertions, always-true literal assertions like Assert.IsTrue(true), self-referential assertions, or un-awaited async assertions. Tests with only trivial checks such as a lone IsNotNull also score lower than tests verifying meaningful values or exceptions.

Avoid it for full suite audits, writing new tests, fixing failing tests, or measuring code coverage. Those tasks belong to suite-audit skills, test-generation agents, editing skills, or coverage-analysis tools respectively.