Eu vou fazer configuração de Claude code com safety hooks, skills personalizadas e Claude MD

Claude Code é poderoso, mas pronto para uso ele pode rodar rm -rf com um espaço extra, instalar pacotes que não existem e consumir seus limites de uso.
I set it up properly for you. I build and maintain claude-skills, an open source collection of 23 Claude Code skills and hooks, tested on Windows, macOS and Linux. I use them every day.
O que eu posso fazer:
- Safety hooks: block destructive commands (with snapshots to undo), stop fake or typo packages, block commits that contain API keys
- Audit your setup: secrets saved in permission rules, risky allow rules, MCP config
- Custom skills for your workflow: deploys, QA, reports, content
- A CLAUDE.md and project rules that Claude actually follows
- Usage audit: see where your limits go and how to cut waste
Everything is plain code you can read, and you keep all of it. No calls needed: you get the files plus a clear written guide.
Message me first with your OS and what you use Claude Code for, and I'll tell you which package fits.