grade-tests

Grade Tests

Grade a curated list of test methods and produce a compact, PR-comment-friendly report: one row per test method with a letter grade, a score band, pseudo-mutation resilience, a one-line note explaining the grade, and a concrete improvement for every grade below A. The skill does not discover tests on its own — the caller (typically a PR automation workflow or a human reviewer holding a specific list) provides the test methods to grade.

Language-specific guidance: Call the test-analysis-extensions skill to discover available extension files, then read the file matching the target codebase's language and framework (e.g., extensions/dotnet.md, extensions/python.md, extensions/typescript.md, extensions/go.md). You MUST read the relevant extension file before scoring assertions or anti-patterns, because assertion APIs and idiomatic patterns differ significantly across frameworks.

Why a Per-Test Grade

Suite-wide audits (test-anti-patterns, assertion-quality, test-smell-detection) produce excellent diagnostic reports, but they are hard to consume as a short PR comment. Reviewers of a PR mostly want to know: for the tests this PR adds or changes, are they good? This skill answers that question with a one-row-per-test verdict that fits in a comment table.

When to Use

• A PR automation workflow needs to post a comment grading the tests introduced or modified in a pull request.
• A reviewer has a specific list of tests (a file, a class, a method list, or a diff hunk) and wants a per-test verdict rather than a suite report.
• A maintainer wants to triage which of N tests in a contribution deserve follow-up improvements.

When Not to Use

• The caller wants a full suite audit or comparative metrics — use test-anti-patterns (pragmatic) or test-smell-detection (formal) and let the test-quality-auditor agent orchestrate.
• The caller wants to write new tests — use code-testing-generator (any language) or writing-mstest-tests (MSTest specifically).
• The caller wants to measure code coverage or CRAP scores — use coverage-analysis or crap-score (.NET only).
• The caller wants to fix issues directly in test code — invoke the appropriate editing skill.
• No specific list of tests is provided. Do not try to grade every test in the workspace; ask the caller for an explicit list or scope.

Inputs

Step 0: Validate the input

Before doing anything else, check that the caller provided one of:

1. An explicit list of test method names, or
2. One or more file paths plus an explicit instruction to grade every test declared in those files (e.g., "grade every test in OrderTests.cs"), or
3. A diff hunk or PR identifier whose changed tests should be graded.

If the request is ambiguous (e.g., "Grade my tests", "Are these tests any good?" with no scope, "Review the test suite"), do not load extensions, do not read files, and do not grade anything. Reply with a short message asking the caller to provide an explicit list / file(s) / diff, and optionally point them at test-quality-auditor agent or test-anti-patterns skill for full-suite analysis. Stop there.

Workflow

Step 1: Detect language and load extension

Identify the target codebase's language and test framework from the file extensions and the test method markers in the provided list. Call the test-analysis-extensions skill and read the matching extension file (e.g., extensions/dotnet.md for MSTest/xUnit/NUnit/TUnit, extensions/python.md for pytest, extensions/typescript.md for Jest/Vitest, extensions/go.md for the standard testing package). If the input contains tests from multiple languages, load each relevant extension and grade each test using its language's conventions.

Step 2: Resolve the test bodies

For each entry in the input list:

1. If the test body is provided inline, use it directly.
2. Otherwise read the file at the given path and locate the method by its fully-qualified name. Capture the full method body, including attributes / decorators / fixtures and any helper code that the test calls.
3. If a method cannot be found, record it as N/A — method not found and continue. Never invent a body to grade.
4. Trace the production entry point called by the test, including every helper and branch relevant to the behavior claimed by the test, even when its current input does not reach that branch. If the production code is not supplied, resolve it from the workspace using symbol references and call sites. Record production code as unavailable only after a reasonable lookup fails.

Step 3: Score each test

Start every test at grade A (score band 90–100), then apply deductions strictly for observable issues in the captured body. Do not deduct for hypothetical concerns (e.g., "could have more negative assertions") unless the production code clearly demands them and the production code is available.

Four sub-dimensions

Compute four sub-grades (each A–F) that together drive the overall grade. Pseudo-mutation may be N/A only when the production code cannot be resolved.

A. Assertion strength

Read the loaded language extension's assertion API list and classify every assertion in the test body. Score from highest to lowest:

Exception tests (Assert.ThrowsException<T>, pytest.raises, expect(fn).toThrow, assertThrows, #[should_panic], Should -Throw, EXPECT_THROW) are complete on their own — do not require additional assertions.

B. Structure & focus

C. Anti-pattern hygiene

Scan against the catalog below. The Anti-pattern sub-grade is computed in two passes and combined deterministically:

1. Hard ceiling pass. Every Critical or High finding sets a maximum sub-grade (F, D, or C as labeled). Take the worst ceiling across all matched Critical/High findings — these do not accumulate (a single F finding caps the sub-grade at F regardless of how many other Critical/High findings are present).
2. Medium-deduction pass. Start from A, then for each Medium finding deduct one sub-grade level (A→B, B→C, C→D, D→F). These do accumulate across findings.

The final Anti-pattern sub-grade is the worse of the two passes (i.e., min(hard_ceiling, A − medium_count)). Low findings never affect the grade — mention them in the note only.

Examples (Critical/High and Medium counts → Anti-pattern sub-grade):

• Zero Critical/High, 1 Medium → B (A − 1)
• Zero Critical/High, 3 Medium → D (A − 3)
• One C-ceiling (e.g., over-mocking), 0 Medium → C
• One C-ceiling, 2 Medium → D (min(C, A − 2 = C) = C, but a third Medium would tip to D)
• One F-finding (e.g., swallowed exception) plus any number of Medium → F

Critical (drop straight to F or D)

• No assertions at all → F (also drives Assertion sub-grade to F)
• Swallowed exceptions: try { … } catch { } (.NET), bare except: pass (Python), try { … } catch (e) {} (JS/TS/Java), defer recover() without re-panic (Go), rescue StandardError with no assertion (Ruby), empty catch (Kotlin/Swift) → F
• Assert-in-catch pattern (Assert.Fail(ex.Message) instead of Assert.ThrowsException) → D
• Always-true literal assertions (Assert.IsTrue(true), assert True, expect(true).toBe(true)) → F (verifies nothing; also drives Assertion sub-grade to F)
• All async assertions are neither awaited nor returned (for example, an Assert.ThrowsAsync task or expect(promise).resolves chain whose result is silently discarded) → F
• Self-referential / tautological assertions on bound values (Assert.AreEqual(x, x), assert dto.name == dto.name) → D
• Commented-out assertions → D

High (drop one or two sub-grades)

• Wall-clock sleep used for synchronization: Thread.Sleep, Task.Delay, time.sleep, setTimeout-based wait, Thread.sleep, time.Sleep, sleep, std::thread::sleep, Start-Sleep, std::this_thread::sleep_for (in a unit test) → D
• Unseeded randomness, wall-clock reads without abstraction (DateTime.Now, datetime.now(), Date.now(), System.currentTimeMillis(), time.Now(), Time.now, Instant::now(), Get-Date, system_clock::now) → D
• Hard-coded environment-dependent paths (C:\…, /tmp/…, network hosts) → D
• Ordering dependency on mutable static / package globals → D
• Broad exception assertion (Assert.ThrowsException<Exception>, pytest.raises(Exception), expect(fn).toThrow(Error) without matcher, #[should_panic] without expected = "…", Should -Throw without -ExpectedMessage, EXPECT_ANY_THROW) → C
• Over-mocking: more mock setup lines than test logic, or verifying exact call sequences instead of outcomes → C
• Implementation coupling: reflection on private members, casting to internal types to access state → C

Medium (drop one sub-grade)

• Poor name: Test1, TestMethod, test, single-word name that says nothing about scenario or expected outcome (judge against the language extension's convention) → drop one sub-grade
• Magic values: unexplained 42, "foo", 0x1234 in arrange/assert without naming or comment → drop one sub-grade
• Giant test (>30 lines covering a single behavior) → drop one sub-grade
• Assertion messages that just repeat the assertion text → drop one sub-grade
• Missing AAA / GWT separation when the test is non-trivial → drop one sub-grade

Low (note only, no deduction)

• Unused setup/teardown hooks; print debugging left in (Console.WriteLine, print, console.log, System.out.println, fmt.Println, puts, dbg!, Write-Host, std::cout); inconsistent naming versus siblings; leftover TODO comments. Mention in the note column but do not deduct.

D. Pseudo-mutation resilience

Evaluate whether this individual test would detect plausible defects in the production behavior it claims to verify. Do not give a test credit because another test kills the mutation.

Call the test-gap-analysis skill to load its canonical mutation catalog and calibration rules, then apply them at this skill's per-test scope:

1. Identify meaningful mutation points on every production path relevant to the behavior claimed by the test, including paths its current input does not exercise:
   • boundary changes (< ↔ <=, index or loop off-by-one);
   • Boolean and logic flips (&& ↔ ||, negated conditions);
   • changed return values (null/None/nil, empty, zero, opposite Boolean);
   • removed exceptions, guards, or error propagation;
   • arithmetic changes (+ ↔ -, sign or increment flips);
   • removed null/None/nil checks or coalescing.
2. Classify each point against this test:
   • Killed — this test's inputs reach the point and its assertions would fail.
   • Survived — this test reaches the point but its assertions would still pass.
   • No coverage — the mutation is part of the behavior named by this test, but the test's inputs do not reach it.
   • Equivalent — the mutation cannot change observable behavior; exclude it.
3. Judge only the behavior promised by the test name and setup. Do not penalize a focused test for unrelated branches that belong in separate tests.
4. Skip trivial code such as simple getters, auto-properties, generated code, and boilerplate. Prefer risk-significant logic over mutation count.

Score the non-equivalent mutation points:

For the report, render mutation resilience as killed/total killed (for example, 3/4 killed), 0/0 (no meaningful points), or N/A. Exclude Equivalent mutations from the total.

Combining sub-grades

Convert sub-grades to numeric points: A=4, B=3, C=2, D=1, F=0.

• Overall score band = weighted average: 0.35 × Assertion + 0.30 × Pseudo-mutation + 0.20 × Anti-pattern + 0.15 × Structure
• If Pseudo-mutation is N/A, omit it and renormalize the remaining weights to total 1. Do not lower a grade because production context is unavailable.
• Map to letter:
  • ≥ 3.5 → A (band 90–100)
  • ≥ 2.8 → B (band 80–89)
  • ≥ 2.0 → C (band 70–79)
  • ≥ 1.2 → D (band 60–69)
  • < 1.2 → F (band 0–59)
• Apply an overall ceiling only when the Anti-pattern catalog classifies an observed finding as Critical: a Critical finding labeled F forces the overall grade to F, and one labeled D caps it at D. Do not otherwise cap the weighted result at the worst sub-grade; the weights must remain meaningful.

Report the letter grade and the score band (not a single 0–100 number). False precision invites bikeshedding; bands keep the conversation focused on the rubric.

Step 4: Build the note and improvement

The note column is one short sentence (target ≤ 120 characters). State the single most important reason for the grade. Examples:

• A (90–100): Clear AAA structure; equality + exception assertions on the public contract.
• B (80–89): Primary contract is protected, but the upper-bound mutation survives.
• C (70–79): Only checks IsNotNull on the result; no value verification.
• D (60–69): Self-referential assertion: round-trip identity verifies plumbing, not transformation.
• F (0–59): No assertions — test executes the method but never verifies anything.

If a test gets A with no notable issues, the note may simply be No issues found. — do not invent weaknesses to justify the grade.

For every test below A, add a How to improve sentence (target ≤ 120 characters) describing the smallest concrete change that addresses the grade-limiting issue:

• Name the assertion to add or strengthen and the observable value to check.
• For a survived mutation, give the input and expected outcome that would kill it, such as Add max+1 input and assert ArgumentOutOfRangeException.
• For structure or hygiene findings, name the specific extraction, split, or deterministic replacement to make.
• Do not use vague advice such as Add more assertions, Improve coverage, or Write better tests.
• Do not recommend unrelated scenarios merely to make a focused test broader.
• For A-grade tests, render —; no improvement is required.

Step 5: Report

Produce two sections.

1. Summary

A short paragraph (2–4 sentences) covering: total tests graded, grade distribution, the most common issue, and the single highest-leverage recommendation. Include the most important survived or uncovered mutation only when one exists. Otherwise state the applicable result without inventing a gap: all meaningful mutations were killed, no meaningful mutation points were found (0/0), or production code could not be resolved (N/A). In the latter two cases, lead with the dominant assertion, hygiene, or structure signal.

2. Per-test table

| Test | Grade | Band | Mutation | Notes | How to improve |
|------|-------|------|----------|-------|----------------|
| `Namespace.ClassName.Test_Method_Condition_Expected` | A | 90–100 | 4/4 killed | Clear AAA; assertions protect the public contract. | — |
| `Namespace.ClassName.Test_UpperBound` | B | 80–89 | 2/3 killed | The inclusive upper-bound mutation survives. | Add `max + 1` input and assert rejection. |
| `Namespace.ClassName.Test_Other` | C | 70–79 | 1/3 killed | Only `IsNotNull`; default-return mutations survive. | Assert the expected value and collection contents. |
| `Namespace.ClassName.Test_Old` | F | 0–59 | 0/2 killed | No assertions; every mutation survives. | Assert the exact result produced by the arranged input. |

Caps and ordering:

• If the table would exceed 50 rows, show all tests graded below B first (worst to best), then a sample of the best tests, and wrap any overflow in a collapsed <details> block.
• Within the same grade, order by file path then by method name for determinism.
• If the diff context is provided, prefix each test name with a (new) or (modified) marker.

If multiple languages are present, produce one table per language and prefix each section with the language name and framework.

Validation

• Every test in the input list appears in the table (or is recorded as N/A — method not found).
• Every grade is justified by at least one observable signal in the captured body — no speculative deductions.
• Production code was resolved where possible; unavailable code is reported as mutation N/A and omitted from scoring.
• Every meaningful mutation point in the behavior claimed by each test is classified Killed / Survived / No coverage / Equivalent.
• Equivalent mutations and unrelated production branches do not affect the mutation sub-grade.
• Mutation resilience is evaluated per test; a mutation killed only by a different test does not receive credit.
• Trivial-assertion tests are flagged only when the only assertion is trivial (a null check before a meaningful assertion is not trivial).
• Exception-only tests are not penalized for low assertion count.
• Mock-call verifications and bare assertion forms count as real assertions of the appropriate category.
• Boolean assertions on meaningful properties (Assert.IsTrue(result.IsValid)) are not classified as always-true; only literal true/false constants are.
• Self-referential assertions are flagged separately from normal equality assertions.
• Idiomatic patterns are not flagged: Go/Rust table-driven sub-tests, pytest bare assert, Go if got != want { t.Errorf(...) }, JS/TS expect(mock).toHaveBeenCalledWith(...).
• Async test pitfalls (un-awaited resolves/rejects/ThrowsAsync, pytest-asyncio without await) drop the Assertion sub-grade to F.
• Every test below A has one concrete, grade-targeted improvement; every A test uses —.
• The summary leads with the highest-leverage observation, not a recap of the table.

Common Pitfalls