Claude Skillforge
nicholashidalgo/claude-skillforge
Biblioteca modular de skills battle-tested para Claude com carregamento sob demanda
Ver no GitHubBiblioteca de production-ready de Claude Skills modulares e reutilizáveis, cobrindo engenharia de dados, escrita profissional e ferramentas de carreira. Cada skill é uma pasta auto-contida com instruções YAML que Claude carrega dinamicamente sob demanda, economizando tokens através de divulgação progressiva.
Instalação
npx convez-ai nicholashidalgo/claude-skillforge1 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
- Estender capacidades do Claude com instruções especializadas reutilizáveis
- Implementar workflows profissionais padronizados (escrita, análise de dados, career coaching)
- Gerenciar contexto eficientemente com skills que carregam apenas quando relevantes
Veredito

Skillforge é uma abordagem estruturada e pragmática para composição de comportamentos no Claude. O foco em production-readiness e economia de tokens via progressive disclosure diferencia de listas ad-hoc, mas carece de documentação abrangente no README para onboarding rápido de novos usuários.
README
A tradução para o português deste README ainda está na fila.
What is this?
Claude Skillforge is a multi-category library of production-ready Claude Skills. It is not a writing-skills repo. It spans foundational operating standards, data engineering, professional writing, and career tools - every skill battle-tested in real workflows, not written as a demo.
Skills are instruction files that teach Claude how to perform specific tasks in a repeatable, high-quality way. They load on demand and cost roughly 100 tokens until activated.
What are Claude Skills?
Claude Skills are specialized instruction folders that Claude dynamically discovers and loads when relevant to a task. Each skill contains a SKILL.md file with YAML frontmatter (name and description for routing) and detailed execution instructions.
Skills use progressive disclosure: Claude scans metadata first (~100 tokens), loads full instructions only when the skill matches (~2-5K tokens), and accesses bundled resources only as needed. Multiple skills can be active simultaneously without overwhelming the context window.
How to install
Claude.ai (Web/Desktop)
- Go to Settings > Capabilities > Skills
- Upload skill folders or point to this repository
Claude Code (CLI)
git clone https://github.com/nicholasjh-work/claude-skillforge.git
# Copy desired skill folders into your project's .claude/skills/ directory
Manual Copy any skill folder into your Claude skills directory. Each folder is self-contained.
Spotlight
Skill of the Week
Watches conversation length and auto-fires a continuation handoff before you hit the token limit. Companion to claude-session-handoff.
Skill of the Month
Detects and removes 38 documented AI writing patterns. Two modes: CLEAN for professional docs, VOICE for creative writing.
Updating the spotlight: edit this section in
README.md. Replace the badgelabel-name-colorand the description paragraph. Keep thefor-the-badgestyle and category color (Core0550ae, Data Engineering0d6e3f, Writingb35900, Career6f42c1).
Core
Foundational operating standards that govern how Claude behaves across all tasks.
Data Engineering
Tools for data quality, SQL, reporting, KPI governance, and analytics engineering.
Writing
Professional and editorial writing tools for business communication, technical documentation, and AI prose cleanup.
Career
Resume optimization, salary negotiation, and job application tools.
Design philosophy
These skills follow two layers of standards:
Standards-backed rules grounded in published engineering guidance (PEP 8, Twelve-Factor App, OWASP, dbt best practices, Microsoft Power BI documentation). See docs/rationale-and-references.md.
House operating conventions that are intentional working preferences for AI-assisted engineering: lead with the answer, sparse comments, no AI-style clutter, fixed output structures. These are not claimed as universal standards. See docs/adr/0001-ai-session-operating-standard.md.
Documentation
| Document | Purpose |
|---|---|
| docs/rationale-and-references.md | Standards-backed justification for skill rules with source citations |
| docs/adr/0001-ai-session-operating-standard.md | Architecture Decision Record explaining why these standards exist |
Contributing
Contributions welcome. See CONTRIBUTING.md for guidelines.
License
MIT License. See LICENSE for details.