Ed3Design Skill Bundles
Ed3Design/ed3design-skill-bundles
56 skills + 5 tools + 6 hooks para engenharia de software com Claude Code
Ver no GitHubBiblioteca de 56 skills, 5 ferramentas Python, 6 hooks e 4 sub-agentes organizados em 6 bundles temáticos para engenharia de software. Cobre otimização de tokens, qualidade de código, planejamento, debugging assíncrono, validação de schema e meta-sistema de skills. Padrões validados empiricamente em TDD e licenciados MIT.
Instalação
npx convez-ai Ed3Design/ed3design-skill-bundles1 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
- Otimizar consumo de tokens em prompts e contexto
- Automatizar revisão de código, linting e hygiene Git
- Debugar concorrência assíncrona e containers
- Validar schemas SQL e prevenir bugs de banco de dados
- Estruturar planejamento de features e decisões arquiteturais
Veredito

Recurso robusto e maduro para engenheiros que trabalham intensivamente com Claude Code. Oferece padrões reais de engenharia estruturados para reutilização, com suporte direto via marketplace e validação automatizada. Ideal para equipes que precisam padronizar workflows de software engineering em larga escala.
README
A tradução para o português deste README ainda está na fila.
ed3design-skill-bundles
Software-engineering-discipline library for Claude Code power users. 56 skills + 5 Python tools + 6 hooks + 4 sub-agents across 6 thematic bundles.
📦 Bundles (Software Engineering)
| Bundle | Skills | Hooks | Agents | Tools | Status |
|---|---|---|---|---|---|
token-savers | 4 | 1 | — | 5 | ✅ v0.2.0 |
code-quality | 21 | 4 | 1 (code-reviewer) | — | ✅ v0.2.0 |
planning-disciplines | 10 | — | 1 (planning-assistant) | — | ✅ v0.2.0 |
async-forensik | 8 | — | 1 (forensik-detective) | — | ✅ v0.2.0 |
schema-discipline | 8 | — | 1 (schema-validator) | — | ✅ v0.2.0 |
skill-system-meta | 5 | 1 | — | — | ✅ v0.2.0 |
Total: 56 skills + 6 hooks + 4 sub-agents + 5 tools — patterns extracted from real-world software engineering practice and structured for reuse (Python/SQL/Git workflows).
🔧 Hardware / Maker Engineering — Separate Repo
Hardware-focused bundles (parametric CAD, FDM/maker workflows, embedded systems) moved to a dedicated sibling repo for clearer audience differentiation:
→ ed3design-engineering-bundles — 8 skills across 3 bundles (cad-design, maker-fdm, embedded-systems). Public, MIT licensed.
For makers, mechanical engineers, embedded hobbyists. Includes the Fusion 360 MCP-Bridge skill for Claude-driven CAD automation and the Victron Cerbo Modbus device-onboarding skill (encodes the non-obvious Enabled=0 default).
🚀 Quickstart
Via Claude Code Marketplace (Recommended)
Add this repo as a marketplace, then install bundles individually:
/plugin marketplace add Ed3Design/ed3design-skill-bundles
/plugin install token-savers@ed3design-skill-bundles
/plugin install code-quality@ed3design-skill-bundles
# ... or any of the 6 bundles
Available plugins after marketplace add:
token-savers— Token optimization (4 skills + 5 tools + 1 hook)code-quality— Code review + Git hygiene (21 skills + 4 hooks + 1 agent)planning-disciplines— Roadmap + decision (10 skills + 1 agent)async-forensik— asyncio + container debugging (8 skills + 1 agent)schema-discipline— SQL/schema bug prevention (8 skills + 1 agent)skill-system-meta— Skill lifecycle (5 skills + 1 hook)
Manual install (no marketplace)
git clone https://github.com/Ed3Design/ed3design-skill-bundles
cd ed3design-skill-bundles
# Install one bundle (e.g. token-savers)
ln -s "$(pwd)/token-savers/skills"/* ~/.claude/skills/
cp token-savers/tools/*.py ~/.claude/tools/
chmod +x ~/.claude/tools/*.py
# Optional dependencies (token-savers):
pip install '.[token-savers]' # Pillow + beautifulsoup4 + certifi (from pyproject.toml)
# or individually:
pip install Pillow beautifulsoup4 certifi
brew install tesseract tesseract-lang poppler # OCR + PDF (system binaries)
💡 Design Philosophy
These bundles emerged from intensive engineering practice — every pattern is empirically hardened in real software engineering sessions, not theoretically designed.
Pattern Discovery Methodology: each skill goes through a TDD-promotion cycle (RED + GREEN sub-agent test) before being added to a bundle. See skill-system-meta/skills/skill-tdd-promotion-workflow/.
Generalization Criterion: a skill is bundled only if it's cross-domain useful — independent of specific projects, business domains, or proprietary infrastructure. Project-specific patterns stay in the original catalog.
📐 Bundle Format
Each bundle follows the Claude Code plugin convention:
<bundle-name>/
├── .claude-plugin/
│ └── plugin.json # Plugin manifest
├── skills/ # SKILL.md per skill (frontmatter + content)
│ └── <skill-name>/SKILL.md
├── hooks/ # Optional: PreToolUse/Stop/UserPromptSubmit hooks
│ ├── hooks.json
│ └── <hook-name>.sh
├── agents/ # Optional: custom sub-agent definitions
│ └── <agent-name>.md
├── tools/ # Optional: Python helper scripts
└── README.md # Bundle description + trigger table
Portable Tool Paths: skills reference tools via ${CLAUDE_PLUGIN_ROOT:-$HOME/.claude}/tools/X.py — works both in the plugin context and locally installed.
🔗 Related Projects
- superpowers by Jesse Vincent — plugin format inspiration + TDD-for-skills pattern
📜 License
MIT.
🙏 Acknowledgements
Methodology draws from the superpowers plugin ecosystem (Jesse Vincent). The TDD-promotion-workflow + post-session-skill-review discipline emerged from sustained practice.