Claude Skills — Kit de Revisão e Scaffolding
FixPortal/claude-skills
Revisão adversarial multi-vendor + scaffolding de projetos em PowerShell
Ver no GitHubConjunto curado de skills reutilizáveis para Claude Code, especializado em revisão adversarial de código (multi-vendor), scaffolding de projetos (.NET, React, testes) e gestão de sessões. Inclui workflows para recap, reflect e auditorias documentadas com padrões house-standard.
Instalação
npx convez-ai FixPortal/claude-skills1 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
Casos de uso
- Revisão cruzada de código (Claude + Sonnet + GPT) para capturar erros não-correlacionados entre vendors
- Scaffolding rápido de projetos .NET, React/Vite, testes xUnit e CI/CD (GitHub Actions)
- Auditoria honesta de setup Claude Code e geração de relatórios de inteligência sobre cobertura e riscos
Veredito

Skill altamente especializada para times de desenvolvimento que adotaram Claude Code e querem automação de revisão multi-vendor e scaffolding padronizado. Exige setup local (.claude/skills) e substituição de placeholders; valor máximo em fluxos de CI/CD e auditorias contínuas de qualidade.
README
A tradução para o português deste README ainda está na fila.
Claude Code Skills
A curated, sanitised subset of the authored Claude Code skills I use day to day, published as a portfolio reference. These are the genuinely reusable ones — scaffolding, review, and session-management workflows — with machine paths, client names, and personal vault locations replaced by placeholders.
These are extracted from a larger private working set. Paths like
~/.claude/...,<vault>,<workdir>, and example values likeyou@example.com/Acmeare placeholders — point them at your own locations before use.
What's here
| Skill | What it does |
|---|---|
adversarial-review | Cross-vendor code review — Claude Opus + Claude Sonnet + a non-Claude model (GPT via the GitHub Copilot CLI), then cross-examination and adjudication. The point is uncorrelated error across vendors. |
scaffold-dotnet | Create / normalise a .NET solution to a house standard (NodaTime boundaries, central package management, canonical .editorconfig). |
scaffold-tests | Scaffold xUnit v3 + NSubstitute + AwesomeAssertions test projects. |
scaffold-ci | GitHub Actions CI for .NET / Vite-React / hybrid repos, plus Dependabot and CodeQL. |
scaffold-frontend | Vite + React + TypeScript scaffolding with ESLint (sonarjs) and Vitest. |
scaffold-minimal | Convert ASP.NET controllers to minimal APIs with OpenAPI + Scalar. |
scaffold-doc | Author structured markdown docs (READMEs, audit reports, runbooks) in a consistent house style. |
review-digest | Mine PAST adversarial-review / reviewer-findings work across a folder of repos into a dated intelligence report — coverage ledger, recurring-theme digest, forward-looking risk ranking, and a hand-off scope brief (commits since each repo's last review + a paste-ready review prompt for another agent). Read-only; runs no reviews. |
recap | "Where did we get to?" — a fast, multi-source, journalled recap of work done and what's next, per repo/branch. |
reflect | An honest, evidence-scored audit of how well a Claude Code setup is actually being used. |
How skills work
Each folder is a skill: a SKILL.md with YAML frontmatter (name, description)
that Claude Code loads on demand when the description matches the task. Drop a
folder into ~/.claude/skills/ (global) or a repo's .claude/skills/ (project)
and it becomes available.
A note on the adversarial-review skill
The whole value is the non-Claude reviewer. A multi-agent panel made only of Claude models is same-vendor self-review and defeats the purpose — the uncorrelated error of a different vendor's model is what catches what one vendor misses. The skill wires Claude Opus + Claude Sonnet + GPT (via the Copilot CLI).
Licence
MIT — see LICENSE.