Test Coverage

STOP - Read before ANY tool call

1. Read {{GUARDRAILS_PATH}}
2. Read _shared/sdd-artifacts/SESSION.md; load session-state for $Cwd
3. If the relevant gate is not approved: STOP - ask user (pt-BR) - do NOT Write/Shell
4. SDD/develop skills: after ONE step/task, STOP session - handoff only
5. This skill body is English; user-facing prompts may be (pt-BR)

Step -1 - Gate check (report in chat before continuing)

Gate check:
[ ] guardrails.mdc read
[ ] SESSION.md read; session-state loaded
[ ] PIPELINE.md read (SDD skills only)
[ ] User confirmed current action (sim)
-> If any unchecked: STOP

Trigger

Invoke when the user asks for: /test-coverage, coverage report, coverage, or when a PLAN step / code-review requires coverage evidence.

Arguments (optional):

Outcome

A structured coverage report in pt-BR with:

• Coverage on new code - line coverage on changed production files vs base branch
• Overall branch coverage - solution-wide line coverage after tests
• Per-file breakdown - each changed production file with line %
• Decision: Pass (>= threshold) or Fail (< threshold) with gap list

Does not modify code unless the user asks for test additions in a follow-up.

Lazy-load

Never by default: do not preload all references/*.md or full guideline packs. Load one section file per Process step — never full reference.md when a section exists (SKILL-REFERENCE-RETRIEVAL.md).

Reference routing

Process

Read references/<section>.md for commands, metrics, and templates — not full reference.md.

Step -1b - Caveman Mode (Full cap)

1. Read {{SDD_ROOT}}/preferences.json (create { "caveman_mode": false, "caveman_level": "full" } if missing).
2. If caveman_mode is false: continue without compression.
3. If true: load {{TOOLKIT_ROOT}}/skills/_shared/caveman/CAVEMAN.md; apply Full participation cap + prefs caveman_level (Lite skills never escalate); show once: [Caveman] Modo ativo (respostas compactas, level={effective}). Digite caveman off para desativar.
4. Honor caveman on|off|status|lite|full|ultra (and stop caveman / normal mode) during the session.
5. Auto-Clarity + never-compress gates/drafts/paths per CAVEMAN.md.

-1. Mode

PIPELINE.md: Agent required for dotnet test and ReportGenerator. In Plan/Ask, explain limitation and list expected paths under TestResults/ after the user switches to Agent.

0. Workspace

Confirm target .NET repository (.sln or *Tests.csproj). If the workspace is agent-dev-toolkit only (no test projects), stop and ask which consumer repo to open. Detect stack; read AGENTS.md / README.md when present.

1. Resolve scope

Base branch: user argument, or main / develop (ask once if ambiguous).

git fetch origin   # when remote comparison is needed
git rev-parse --abbrev-ref HEAD
git diff <base>...HEAD --name-only -- "*.cs"

Filter to production changed files per references/exclusions.md. Record the list for per-file metrics. Large repos: see references/scoped-discovery.md.

2. Prerequisites check

Before running tests, verify per references/prerequisites.md. If coverlet.collector is missing, stop with install instructions - do not fail silently.

3. Collect coverage (mandatory ReportGenerator)

1. Run dotnet test with Coverlet -> TestResults/**/coverage.cobertura.xml per references/commands.md.
2. Always run ReportGenerator -> TestResults/CoverageReport/ (Summary.txt, index.html, Cobertura). Do not finish with XML only. Use scoped test project when the repo is large or user provided a path.

4. Compute metrics

Parse Cobertura / ReportGenerator output per references/metrics.md (new code / overall branch / per-file). Exclude migrations, generated code, and test projects from new code denominator (references/exclusions.md).

5. Evaluate threshold

6. Report (chat + on-disk artifacts)

Use references/report-template.md. Cite paths from references/artifacts.md. Required in chat: Summary.txt / index.html / cobertura paths, Resumo table, commands, limitations, approval block (Pass) or gaps (Fail). Do not claim Pass if ReportGenerator output or Cobertura files are missing. Integrations: references/integrations.md. Troubleshooting: references/troubleshooting.md.

7. Handoff

Must not

• Require SonarLint, Visual Studio, SonarQube login, or corporate pipeline APIs
• Auto-commit, auto-push, or add tests without user request
• Claim Pass when tests did not run, coverlet output is missing, or ReportGenerator was skipped
• Deliver coverage only in chat without citing on-disk paths under TestResults/
• Count EF migrations, *.g.cs, or *.Designer.cs in new-code denominator
• Block merge by itself - gate is informational unless PRD/PLAN/code-review applies threshold