csharp-test-coverage

name	csharp-test-coverage
description	Run C# tests with code coverage and report per-class/method line and branch coverage with uncovered line numbers. Accepts an optional test filter and optional class filter.
allowed-tools	["Bash(dotnet test:*)","Bash(reportgenerator:*)","Bash(node:*)","Bash(rm:*)","Read"]

Run C# tests with coverlet code coverage collection and produce a per-class coverage summary including uncovered line numbers for the classes you care about.

Arguments

• $ARGUMENTS — An optional string with two parts (both optional):

  • Test filter: a dotnet test --filter expression (e.g. FullyQualifiedName~ProjectControllerTests)
  • Class filter: prefixed with --classes followed by comma-separated partial class names to include in the report (e.g. --classes Projects,ProjectController)

  If no test filter is provided, all tests are run. If no class filter is provided, all classes with >0% coverage are shown.

Examples

/csharp-test-coverage FullyQualifiedName~ProjectControllerHttpTests --classes Projects,ProjectController
/csharp-test-coverage --classes Agreements,AgreementController
/csharp-test-coverage FullyQualifiedName~AuthorizationTests
/csharp-test-coverage

Steps

1. Parse arguments from $ARGUMENTS:

   • Split on --classes to extract the test filter (before) and class name filters (after, comma-separated).
   • If no --classes flag, treat the entire argument as the test filter.

2. Run tests with coverlet to collect coverage in cobertura format:

   dotnet test WADNR.API.Tests/WADNR.API.Tests.csproj [--filter "<test-filter>"] -p:CollectCoverage=true -p:CoverletOutputFormat=cobertura -p:CoverletOutput=./coverage-results/ --no-restore
   

   • If a test filter was provided, include --filter "<test-filter>".
   • If tests fail, report the failure and stop.

3. Extract per-class coverage with uncovered lines using the helper script:

   node .claude/skills/csharp-test-coverage/parse-coverage.js WADNR.API.Tests/coverage-results/coverage.cobertura.xml [classFilter1,classFilter2,...]
   

   • If class name filters were provided, pass them as a comma-separated second argument.
   • The script outputs one JSON object per line with: class, file, lineCoverage, branchCoverage, uncoveredLines.
   • The uncoveredLines field contains line ranges (e.g. "18-20, 34-36") referring to the source file.

4. Present results as a markdown table with columns: Class, File, Line%, Branch%, Uncovered Lines.

5. Build a prioritized action plan toward 100% coverage. For every class with < 100% line or branch coverage:

   a. Read uncovered lines — Use the Read tool to read the specific uncovered line ranges from each source file.

   b. Categorize each gap by difficulty:

   Difficulty	Criteria
   Quick win	≤ 5 uncovered lines, simple null checks / early returns / trivial branches, no new test infrastructure needed
   Standard	6–20 uncovered lines, CRUD paths / validation logic / moderate branches, needs a new test method but no new mocks or fixtures
   Complex	> 20 uncovered lines, multi-step flows / external-service interactions / spatial-geometry code / authorization edge cases, needs new mocks or test infrastructure

   c. Assess priority based on what the code does:

   Priority	Applies to
   High	Core business logic, security/authorization paths, data mutation (create/update/delete)
   Medium	Read paths, lookup/reference data, grid/list projections
   Low	Logging, defensive null guards on already-validated internal data, infrastructure plumbing — still list these but note why they are low priority

   d. Output an action-plan table sorted by priority (High first), then difficulty (Quick win first within same priority):

   | # | Class | Gap Description | Lines | Priority | Difficulty | Suggested Test |
   |---|-------|----------------|-------|----------|------------|----------------|
   | 1 | ProjectController | Delete — missing 403 and 404 branches | 82-95 | High | Quick win | Add Delete_Returns403_WhenUnauthorized and _Returns404 |
   | 2 | Agreements | CreateAsync — null FundSource branch | 44-51 | High | Standard | Seed Agreement without FundSource, assert validation error |
   

   Every uncovered line range MUST appear in the table — never silently skip gaps.

   e. Summarize at the bottom:

   • Total gaps by priority × difficulty (e.g. "3 High/Quick-win, 2 High/Standard, 1 Medium/Complex")
   • Recommended next batch: pick the top items (up to ~5) that maximize coverage gain with minimum effort — label these as "do next"
   • Remaining coverage gap as a percentage: what line% and branch% would reaching if the "do next" batch is completed

6. Clean up temporary coverage files:

   rm -rf WADNR.API.Tests/coverage-results/
   

Notes

• The test project uses coverlet.msbuild (a dependency in WADNR.API.Tests.csproj).
• reportgenerator is installed as a global dotnet tool.
• Coverage files are written to WADNR.API.Tests/coverage-results/ and cleaned up after reporting.
• The --no-restore flag speeds up the test run by skipping NuGet restore.
• The helper script parse-coverage.js lives alongside this SKILL.md and parses cobertura XML using Node.js.