dotclaude
dopiotrek/dotclaude
Framework agentic para Claude Code com hooks, agentes e skills
Ver no GitHubFramework configurável para Claude Code que fornece guardrails automatizados, agentes especializados, skills reutilizáveis e padrões sensatos para desenvolvimento assistido por IA. Inclui 13 hooks de validação/segurança, 9 agentes de domínio específico e configurações compartilháveis.
Instalação
npx convez-ai dopiotrek/dotclaude1 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
- Automatizar validações de segurança e formatação de código
- Executar tarefas especializadas (frontend, backend, debugging)
- Compartilhar configurações e padrões de desenvolvimento em equipes
- Monitorar e auditar operações do Claude Code
Veredito

Solução madura e bem estruturada para quem já usa Claude Code profissionalmente. Excelente para equipes que querem padronização, segurança e reutilização de configurações, mas requer familiaridade com hooks Python e fluxos de desenvolvimento ágil.
README
A tradução para o português deste README ainda está na fila.
dotclaude
Claude Code agentic framework - hooks, agents, skills, and configuration
A comprehensive, shareable configuration framework for Claude Code that provides automated guardrails, specialized agents, reusable skills, and sensible defaults for productive AI-assisted development.
What's Included
| Component | Count | Description |
|---|---|---|
| Hooks | 13 | Automated guardrails for security, formatting, and validation |
| Agents | 9 | Specialized task handlers (frontend, backend, debugging, etc.) |
| Skills | 4 | Reusable skill definitions with reference materials |
| Settings | 1 | Comprehensive permission rules and hook configuration |
| CLAUDE.md | 1 | Global preferences and instructions |
Quick Start
# Clone the repository
git clone https://github.com/dopiotrek/dotclaude.git ~/repos/dotclaude
# Run the installer (creates symlinks)
cd ~/repos/dotclaude
./install.sh
# Start a new Claude Code session
claude
The installer creates symlinks from ~/.claude to this repository, making it the single source of truth. Any changes you make here automatically apply to Claude Code.
Installation Options
./install.sh # Default: symlink mode (recommended)
./install.sh --copy # Copy files instead of symlinks
./install.sh --uninstall # Remove symlinks and optionally restore backup
Hooks
Hooks are Python scripts that run before/after Claude Code tool invocations, providing automated guardrails:
Security & Safety
- no-secrets.py - Blocks hardcoded API keys, tokens, and passwords
- permission-auto-approve.py - Auto-approves safe read-only operations
- bash-command-validator.py - Enforces use of modern CLI tools (rg over grep)
Code Quality
- auto-format.py - Auto-formats code after edits (Prettier, Black, etc.)
- import-path-validator.py - Enforces import conventions ($lib, package imports)
- sveltekit-route-validator.py - Validates SvelteKit routing conventions
- sveltekit-perf-guard.py - Monitors bundle size and performance anti-patterns
Database & Migrations
- drizzle-migration-guard.py - Warns about destructive schema changes
- supabase-rls-reminder.py - Reminds to add Row Level Security policies
Utilities
- command-logger.py - Logs all Claude Code operations for audit
- dependency-audit.py - Runs security audits when dependencies change
- stop-verification.py - Runs type checks when Claude completes work
- web-search-enhancer.py - Adds current year to tech documentation searches
See hooks/README.md for detailed documentation.
Agents
Specialized agents handle complex, domain-specific tasks:
| Agent | Model | Purpose |
|---|---|---|
| frontend-engineer | sonnet | Svelte 5 components, shadcn-svelte, responsive design |
| backend-engineer | sonnet | SvelteKit server-side, load functions, form actions |
| superforms-expert | sonnet | sveltekit-superforms + Zod validation |
| debug-specialist | sonnet | Error diagnosis, test failures, performance issues |
| code-reviewer | opus | Comprehensive security and quality review |
| discovery-agent | sonnet | Requirements gathering and feature specification |
| vercel-deployment-expert | sonnet | Vercel deployment and configuration |
| seo-expert | sonnet | Content structure and SEO optimization |
| mobile-ui-designer | sonnet | Mobile-first UI design with TailwindCSS |
See agents/README.md for detailed documentation.
Skills
Skills are reusable task definitions with supporting reference materials:
- clean-comments - Remove unnecessary code comments
- code-reviewer - Code review with checklists and analysis scripts
- frontend-design - UI constraints for consistent design
- svelte-component-architecture - Component decomposition guidelines
See skills/README.md for detailed documentation.
Configuration
CLAUDE.md
Global preferences and instructions that apply to all Claude Code sessions:
# Global Claude Preferences
## About Me
- Primary stack: SvelteKit, TypeScript, Supabase
- All projects use pnpm (never npm or yarn)
## Things to Always Do
- Run type checks before considering work complete
- Preserve existing code style in files being edited
- Use existing patterns from the codebase
...
Settings
The settings/settings.template.json contains:
- Permissions - Fine-grained allow/deny/ask rules
- Hooks - Hook configuration for each event type
- Model - Default model preference
- Plugins - Enabled marketplace plugins
Customization
Adding Your Own Hooks
- Create a Python script in
hooks/:
#!/usr/bin/env python3
import json
import sys
def main():
input_data = json.load(sys.stdin)
tool_name = input_data.get("tool_name", "")
# Your logic here
sys.exit(0) # 0=allow, 1=warn, 2=block
if __name__ == "__main__":
main()
- Add the hook to
settings/settings.template.json - Re-run
./install.shto regenerate settings
Adding Your Own Agents
Create a markdown file in agents/ with YAML frontmatter:
---
name: my-agent
description: When and how to use this agent
model: sonnet
color: blue
---
You are an expert in...
Modifying Skills
Skills in skills/ follow the structure:
skill-name/
├── SKILL.md # Main skill definition
├── references/ # Supporting documentation
└── scripts/ # Automation scripts (optional)
Project Structure
dotclaude/
├── README.md # This file
├── CLAUDE.md # Global preferences
├── install.sh # Installation script
├── settings/
│ └── settings.template.json # Settings with $HOME placeholders
├── hooks/
│ ├── README.md
│ └── *.py # 13 hook scripts
├── agents/
│ ├── README.md
│ └── *.md # 9 agent definitions
├── skills/
│ ├── README.md
│ └── */ # 4 skill directories
└── docs/
└── *.md # Additional documentation
Requirements
- Claude Code CLI
- Python 3.8+ (for hooks)
- Bash (for installation)
Optional Dependencies
Some hooks have optional dependencies for full functionality:
prettier- For auto-formatting JS/TS/Svelteblack- For auto-formatting Pythonterminal-notifier(macOS) - For notifications
License
MIT License - Feel free to use, modify, and share.
Contributing
- Fork the repository
- Create a feature branch
- Make your changes
- Submit a pull request
Ideas for contributions:
- New hooks for additional validations
- Agents for other frameworks/languages
- Skills for common development tasks
- Documentation improvements