Claude Code Toolkit
chrisselig/claude-code-toolkit
Skills e commands reutilizáveis para Python, Git e dashboards
Ver no GitHubColeção de skills e commands reutilizáveis para Claude Code que automatizam workflows comuns em engenharia de dados, Git/GitHub e desenvolvimento de dashboards. Inclui scaffolding de projetos, profiling de dados, deploys Streamlit, automação de PRs e linting.
Instalação
npx convez-ai chrisselig/claude-code-toolkit1 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 workflows completos de pull request (lint, test, commit, push, PR)
- Perfilar e validar qualidade de dados em CSV/Parquet/BD
- Scaffolding e setup de projetos Python com ferramentas padrão
- Deploy e debug de dashboards Streamlit
- Exploração e geração de clientes tipados para APIs
- Limpeza e proteção de branches Git
Veredito

Toolkit prático que reduz fricção em tarefas repetitivas de desenvolvimento. Bem estruturado com skills multi-step e commands focados, mas exige familiaridade com Claude Code e setup manual de symlinking.
README
A tradução para o português deste README ainda está na fila.
Claude Code Toolkit
A collection of reusable Claude Code skills and commands for Python data engineering, Git/GitHub workflows, and dashboard development.
These are the global skills and commands I use across all my projects. They live in ~/.claude/skills/ and ~/.claude/commands/ and are available in every Claude Code session regardless of which project I'm in.
Quick Start
# Clone into your home directory
git clone git@github.com:chrisselig/claude-code-toolkit.git ~/claude-code-toolkit
# Symlink into Claude Code's config
ln -sf ~/claude-code-toolkit/skills/* ~/.claude/skills/
ln -sf ~/claude-code-toolkit/commands/* ~/.claude/commands/
Or copy individual files:
# Just want the PR skill?
cp -r ~/claude-code-toolkit/skills/pr ~/.claude/skills/
What's Included
Skills (interactive workflows)
Skills are multi-step workflows triggered with /skill-name. They guide Claude through a structured process.
| Skill | Trigger | Description |
|---|---|---|
| PR | /pr | Lint, test, branch, commit, push, create PR — the full workflow |
| Data Profile | /data-profile | Profile a CSV/Parquet/DB table for data quality issues |
| API Explore | /api-explore | Hit an API, inspect the response, generate a typed Python client |
| ETL Scaffold | /etl-scaffold | Create a standard Extract-Transform-Load pipeline project |
| Deploy Streamlit | /deploy-streamlit | Deploy or debug a Streamlit dashboard |
| Lint Fix | /lint-fix | Auto-detect linter and fix all errors autonomously |
| New Project | /new-project | Scaffold a new Python project with standard tooling |
| Protect Main | /protect-main | Check and apply branch protection rules |
| Review PR | /review-pr | Checkout, review code, run tests, provide feedback on a PR |
| Cleanup Branches | /cleanup-branches | Delete all merged local and remote branches |
| Data Visualization | /visualization | Create clear, honest charts following Tufte and Few principles |
Commands (quick actions)
Commands are triggered with /command-name and run a focused task.
| Command | Trigger | Description |
|---|---|---|
| Status | /status | Git status, recent commits, open PRs, lint/test health |
| TODO | /todo | Scan for roadmap files and inline TODOs |
| Test | /test | Run test suite, diagnose and fix failures |
| Coverage | /coverage | Test coverage report with untested code highlighted |
| Dependencies | /deps | Audit for outdated, vulnerable, or missing packages |
| Release | /release | Semver tag + GitHub release with auto-generated notes |
| DB Schema | /db-schema | Inspect, compare, or diff database schemas |
| Cron | /cron | List, add, validate, or remove cron jobs |
| Env Setup | /env-setup | Set up or verify a Python dev environment |
My Workflow
I build Python data projects that follow this pattern:
Free API / CSV / Excel
│
▼
┌─────────┐
│ Extract │ httpx, pandas, polars
└────┬─────┘
│
┌────▼─────┐
│ Transform │ polars, Pydantic validation
└────┬─────┘
│
┌────▼─────┐
│ Load │ MotherDuck, Turso, SQLite
└────┬─────┘
│
┌────▼──────┐
│ Dashboard │ Streamlit, Shiny
└────────────┘
These skills and commands are designed around this workflow. They assume:
- Python 3.12+ with conda or venv
- Git + GitHub with conventional commits and PRs
- ruff for linting (falls back to flake8)
- pytest for testing
- Pydantic for data validation
- polars preferred over pandas for large data
- Loguru for logging
Skill Anatomy
Each skill lives in its own directory with a SKILL.md file:
skills/
└── data-profile/
└── SKILL.md
A good skill file has:
- Title — what it does in 3-5 words
- Steps — numbered, specific actions for Claude to follow
- Good/Bad examples — concrete code showing the right and wrong way
- Notes — edge cases, preferences, gotchas
Example structure:
# Skill Name
Brief description.
## Steps
1. First step
2. Second step
3. ...
## Examples
### Scenario Name
**BAD** — why this is wrong:
\```python
bad_code()
\```
**GOOD** — why this is right:
\```python
good_code()
\```
## Notes
- Important caveat
- Edge case to handle
Contributing
This is a personal toolkit, but if you find it useful:
- Fork it
- Add your own skills/commands
- Submit a PR if you think others would benefit
License
MIT