Skills Claude Code by J Scott Chapman
jscottchapman/skills
Skills profissionais Claude Code: dev orientado por tickets, limpeza de personas, design de serviços
Ver no GitHubColeção de skills Python para Claude Code desenvolvidas e benchmarkadas com workflow de skill-creator. Inclui harnesses de desenvolvimento orientado por tickets, ferramentas de limpeza de personas em agentes IA e gerador de ideias para design de serviços. Cada skill é auto-contida com testes, evals e empacotamento pronto para instalar.
Instalação
npx convez-ai jscottchapman/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
- Automação de fluxo ticket-to-PR com test-first workflow
- Auditoria e remoção de personas escondidas em codebases de agentes IA
- Geração de ideias inovadoras para design de customer journeys
Veredito

Conjunto maduro e bem-benchmarkado de skills Python. O destaque é o ticket-dev-harness que automatiza o ciclo plan→tests→code→eval de forma agnóstica ao tracker (Linear, GitHub, Jira). Útil para times que querem rigor e reprodutibilidade em desenvolvimento orientado por issues.
README
A tradução para o português deste README ainda está na fila.
Skills
A collection of Claude Code skills by J Scott Chapman, each developed and benchmarked with the skill-creator workflow (draft, run evals with and without the skill, review, iterate).
Skills in this repo
ticket-dev-harness
A tracker-agnostic ticket-to-PR development harness. Runs a test-first loop (plan, then tests, then implementation, then evaluation), built GAN-style so the tests are written before the code, and right-sized so a one-line fix gets a one-line plan while a big feature gets full rigor. Auto-detects the issue tracker (Linear, GitHub Issues, or Jira), test framework, package manager, and target branch from the repo. The skill lives in ticket-dev-harness/ticket-driven-dev-harness/.
persona-exorcist
Hunts AI agent personas across a project and traces every hidden channel (hooks, memory, CLAUDE.md, sub-agents, settings) that keeps re-invoking a persona you already deleted. Use it when a "killed" persona keeps coming back, or to inventory and safely scrub persona definitions across a codebase, including the user-level global config.
magical-service-design
Generates bold, memorable ideas for designing customer services and experiences, the kind that make people feel something unforgettable. Use it when designing or reimagining a service, customer journey, or onboarding. The skill lives in magical-service-design/skill/magical-service-design/.
Layout
Each top-level folder is a self-contained skill project: the installable skill folder (the one holding SKILL.md), its evals, and a packaged .skill file where one exists. Eval run workspaces are gitignored because they are regenerable and some contain nested git repos.
Installing a skill
Copy or symlink the folder that contains SKILL.md into a skills directory Claude Code reads (for example ~/.claude/skills/), or install the packaged .skill file.