convez.ai
SkillDevOps & InfraestruturaIntermediário
mtschoen

Manutenção de Cobertura 100%

mtschoen/skills-maintaining-full-coverage

Gate de conclusão que força 100% de cobertura de testes com relatório rastreável

Ver no GitHub
InicialPluginScore · de 0 a 100

Skill do Claude Code que bloqueia a conclusão de tarefas até atingir 100% de cobertura de testes, com rastreamento via arquivo de relatório versionado. Implementa uma escada de escalação (testes simples → testes complexos → exceções documentadas) para garantir qualidade de código sem concessões.

Instalação

$npx convez-ai mtschoen/skills-maintaining-full-coverage

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
0
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 1 semana
Versão
—
Issues abertas
0
Contribuidores
2

Casos de uso

  • Garantir qualidade mínima em projetos críticos antes de merge
  • Rastrear regressões de cobertura via git diff automático
  • Treinar agentes a escrever testes abrangentes (mocks, casos extremos)
  • Documentar exceções de cobertura com aprovação explícita

Veredito

Skill bem estruturada com metodologia TDD sólida e 10 cenários de pressão validados. Respeita rigor técnico (100% sem arredondamentos) e oferece escada clara de escalação. Recomendado para equipes que priorizam qualidade e rastreabilidade de métricas de código.

README

A tradução para o português deste README ainda está na fila.

maintaining-full-coverage

A Claude Code skill that gates task completion on 100% test coverage, with a checked-in report file for regression tracking.

What it does

When loaded, this skill enforces three things:

  1. Coverage must be 100% before you claim done. No rounding, no "close enough."
  2. A checked-in report file tracks coverage over time. git diff TEST-REPORT.md catches regressions instantly.
  3. A strict escalation ladder when coverage drops: write tests, heroic testing, ask the human, framework exclusions (with approval), documented exceptions (last resort).

The skill layers on top of test-driven-development (write tests first) and verification-before-completion (prove tests pass). This skill closes the loop on the metric.

Install

Copy SKILL.md to your Claude Code skills directory:

mkdir -p ~/.claude/skills/maintaining-full-coverage
cp SKILL.md ~/.claude/skills/maintaining-full-coverage/SKILL.md

The skill will appear in your /skills list and trigger automatically when completing work in projects that track coverage.

Report file format

The skill expects projects to maintain a checked-in coverage report. Minimal format:

myproject test report — 2026-04-04T12:00:00-07:00
═══════════════════════════════════════════════════

Status:   PASS
Tests:    365 total
Git:      a4f2c91 (add-webhook-support)
Coverage: 1203/1203 statements (100%)
          0 lines uncovered
          1 exclusion annotation

Projects declare the command that generates this file and its location in their CLAUDE.md.

The escalation ladder

When coverage is below 100%:

  1. Write tests -- most uncovered lines are straightforwardly testable
  2. Heroic testing -- mock OS calls, simulate errors, interactive tests for UAC prompts
  3. Ask the human -- they may know a trick, or the code is dead and should be deleted
  4. Framework exclusions -- pragma: no cover, istanbul ignore -- only with human approval
  5. Documented exceptions -- absolute last resort, becomes the new baseline

How it was tested

This skill was developed using the TDD-for-skills methodology: write pressure scenarios, observe baseline agent behavior without the skill, write the skill to fix the gaps, verify compliance.

10 pressure scenarios covering:

ScenarioTests
Close Enough (98.6%)Escalation order, dead code awareness
Platform Code (Linux on Windows)OS mocking, report as first-class artifact
Pragma Shortcut (DB error)Error path simulation, pragma discipline
Batch Testing (no tests at 80%)Development nudge, branch awareness
Documented Exception (FFI)Baseline update, CI adjustment
Elevated/Interactive (UAC)Interactive tests with instructional dialogs
CI Baseline Regression (PR review)CI rejection policy, deadline pressure
Browser/Integration (SPA)Puppeteer, UI audit scripts, multi-suite report
Startup/Shutdown (daemon)Mock init deps, trigger teardown explicitly
Hollow Coverage (code review)Tests that cover without testing behavior

15/15 skill sections covered by at least one scenario. See AUDIT.md for the full coverage matrix.

All RED-GREEN comparisons are in docs/evidence/.

Repo structure

SKILL.md                -- the skill (copy this to install)
AUDIT.md                -- skill coverage audit (15/15 sections)
docs/evidence/          -- pressure scenarios, baseline and GREEN-phase results

License

MIT

Ver o original no GitHub

Bom para

Relacionados