claude-skills-library
frankxai/claude-skills-library
Professional-grade skills for Claude Code & Claude AI - Transform Claude into a domain expert in seconds
Instalação
npx convez-ai frankxai/claude-skills-library1 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
README
A tradução para o português deste README ainda está na fila.
🧠 Claude Skills Library
107 production-grade Agent Skills for Claude Code, Claude.ai & every major agentic runtime
Turn your AI agent into a domain expert. Each skill is a self-contained, spec-compliant
SKILL.md— engineered with real frameworks, current best practices, and opinionated guidance, not 500 words of filler. Free. MIT-licensed. Installable in one command.
⚡ Quick start · 🌐 Browse online · 📚 Full catalog · 🗂️ Categories · ❓ FAQ · 🤝 Contribute
⚡ Quick start
Option A — install as a Claude Code plugin (recommended). One command, auto-updating:
/plugin marketplace add frankxai/claude-skills-library
/plugin install claude-skills-library@claude-skills-library
Skills are now discovered automatically and activate when relevant.
Option B — clone and copy the skills you want:
git clone https://github.com/frankxai/claude-skills-library.git ~/claude-skills-library
mkdir -p ~/.claude/skills
# one skill…
cp -r ~/claude-skills-library/free-skills/mcp-architecture ~/.claude/skills/
# …or all of them
cp -r ~/claude-skills-library/free-skills/* ~/.claude/skills/
Some skills are namespaced (e.g.
free-skills/anthropic/pdf). Copy the folder that directly contains theSKILL.mdyou want. Restart Claude Code and the skills are discovered.
Then just work. Skills trigger on their own, or invoke one explicitly:
"Use the greek-philosopher skill to examine my career decision."
"Design an MCP server with the mcp-architecture skill."
"Review this PR with the github-code-review skill."
"Write a Suno prompt for a cinematic trailer track."
🚀 What is this?
A curated, open-source library of Agent Skills — the portable, model-invoked capability format from Anthropic. Drop a skill into your agent and it gains expert knowledge in a domain: it loads automatically when relevant and stays out of the way when it isn't.
Every skill here:
- ✅ Is a self-contained folder with a
SKILL.mdand optionalreferences/,scripts/,assets/. - ✅ Has spec-compliant frontmatter (
name+description) — enforced byscripts/validate_skills.py. - ✅ Is progressive-disclosure friendly: a tight description for routing, deep guidance loaded only on use.
- ✅ Carries a trigger-rich description so the right skill activates from a library of 100+.
107 skills across AI agents, MCP & SDKs, frontend, Oracle/cloud, content & creative production, engineering workflow, and personal performance. All free. MIT-licensed.
⭐ Featured skills
A small taste — see the full catalog for all 107.
| Skill | Why you'd reach for it | Category |
|---|---|---|
nextjs-expert | App Router, server/client boundaries, caching, and the production gotchas that bite | Web & Frontend |
mcp-architecture | Design an MCP server's resources, tools, prompts, and security from first principles | MCP & SDKs |
oracle-database-expert | Oracle 23ai, Autonomous DB, AI Vector Search, SQL/PLSQL tuning, HA | Oracle & Cloud |
github-code-review | Turn a raw diff into a correctness/security/style review with actionable comments | Engineering |
suno-ai-mastery | Prompt-engineer commercial-quality music with Suno v4.5+ | Creative |
ui-ux-design-expert | Design systems, interaction patterns, and WCAG 2.2 accessibility audits | Web & Frontend |
langgraph-patterns | Graph orchestration, state machines, and human-in-the-loop for agent workflows | AI Frameworks |
greek-philosopher | Socratic questioning and Stoic perspective on a hard decision | Mind & Body |
🗂️ Skill categories
| Category | Count | Examples |
|---|---|---|
| 🤖 AI Agents & Orchestration | 15 | agentic-orchestration, swarm-orchestration, model-routing, reasoningbank-intelligence |
| 🔌 AI Frameworks, MCP & SDKs | 13 | mcp-architecture, mcp-builder, openai-agentkit, claude-sdk, langgraph-patterns |
| ☁️ Oracle & Cloud | 8 | oracle-ai-architect, oracle-database-expert, oci-services-expert, ai-architecture |
| 🎨 Web, Frontend & Animation | 14 | nextjs-expert, ui-ux-design-expert, tailwind, gsap, three, framer-expert |
| 🛠️ Engineering Workflow & GitHub | 9 | github-code-review, performance-analysis, verification-quality, hooks-automation |
| ✍️ Content, Writing & Brand | 14 | brand-voice, book-publishing, social-media-strategy, creator-productivity |
| 🎬 Creative & Media Production | 22 | suno-ai-mastery, video-production-workflow, hyperframes-media, higgsfield-soul-id |
| 🧭 Mind, Body & Philosophy | 5 | greek-philosopher, spartan-warrior, gym-training-expert, health-nutrition-expert |
| 📄 Documents & Productivity | 6 | pdf, docx, pptx, xlsx, product-management-expert, webapp-testing |
| 🧩 Meta & Library | 1 | contribute-catalog |
Counts are generated from the repo by
scripts/generate_catalog.py. See the full catalog for every skill with its description.
🆚 Skills vs. MCP vs. prompts
They solve different problems and compose well together.
| A prompt | An MCP server | An Agent Skill | |
|---|---|---|---|
| What it is | Text you paste | A connection to tools/data | A model-invoked capability |
| Gives the agent | Instructions, once | New actions (APIs, files, DBs) | Expert knowledge + workflow |
| Activation | Manual, every time | Always on once connected | Auto-loads when relevant |
| Token cost | Pays every turn | Tool schemas always present | ~Free until triggered (progressive disclosure) |
| Portability | Locked to one chat | Server you run | A versioned folder, shared across runtimes |
Use a skill to teach the agent how to do something well; use MCP to give it access to a system; use a prompt for a one-off. Many skills here pair naturally with an MCP server.
🌐 Works with every runtime
Agent Skills are a portable format. This library documents import paths for the major agentic runtimes:
| Runtime | Maturity | Guide |
|---|---|---|
| Claude Code | ✅ Native | runtimes/claude-code.md |
Antigravity (agy) | ✅ Substrate-compatible | runtimes/antigravity.md |
| OpenCode | ✅ Adapter pattern | runtimes/opencode.md |
| Codex CLI | 🟡 Per-skill MCP wrap | runtimes/codex.md |
| Gemini CLI | 🟡 Per-skill template | runtimes/gemini-cli.md |
| Cursor | 🟡 Per-skill convert | runtimes/cursor.md |
See MULTI_RUNTIME.md for the universal-MCP rationale and the full strategy.
🧱 Anatomy of a skill
free-skills/<skill-name>/
├── SKILL.md # required — frontmatter + instructions
├── references/ # optional — deep docs loaded on demand
├── scripts/ # optional — executable helpers the agent runs
└── assets/ # optional — templates, fonts, icons used in output
Every SKILL.md opens with two-field frontmatter:
---
name: mcp-architecture # lowercase, hyphenated, ≤ 64 chars
description: Design and implement MCP servers... Use when architecting an MCP server...
---
The description is the most important line you'll write — it's what the model uses to decide
when to load the skill. State what it does and when to use it, with trigger keywords.
📐 Authoring standard: spec/README.md · Start here: template/SKILL.md
✅ Quality & validation
A single zero-dependency validator enforces the standard across every SKILL.md:
python3 scripts/validate_skills.py
# OK: 107 skill file(s) are spec-compliant.
python3 scripts/generate_catalog.py # regenerate docs/CATALOG.md
python3 scripts/generate_catalog.py --check # CI-friendly drift check
python3 scripts/check_links.py # verify no broken internal links
All three run automatically on every push and pull request via GitHub Actions.
It verifies that every skill has well-formed UTF-8 frontmatter, a spec-valid name, a non-empty
description (≤ 1024 chars), and the exact filename SKILL.md; it also warns on oversized bodies
and reserved-word names. Run it before opening a PR.
The browsable catalog at frankxai.github.io/claude-skills-library
is generated by scripts/generate_site.py into docs/index.html and served by
GitHub Pages (Settings → Pages → Deploy from branch → main / /docs).
❓ FAQ
Does loading 107 skills bloat my context?
No. Only each skill's name + description is preloaded for routing; the full body loads only when
a skill triggers, and references/ load only when read. That's the whole point of progressive disclosure.
How do skills get picked from a library this large?
By the description. Every skill here states what it does and when to use it with explicit trigger
keywords, so the model selects the right one. If a skill mis-triggers, sharpen its description — PRs welcome.
Do I have to install all of them?
No. Install the plugin for everything, or copy just the folders you want into ~/.claude/skills/.
Is this affiliated with Anthropic?
No. It's an independent, MIT-licensed community library built on Anthropic's open Agent Skills format.
Some skills under free-skills/anthropic/ are upstream reference skills that retain their original licenses.
Can I use these outside Claude Code?
Yes — see runtimes/ for Antigravity, OpenCode, Codex, Gemini CLI, and Cursor.
🤝 Contributing
New skills, deeper examples, fixes, and better docs are all welcome.
- Read
CONTRIBUTING.mdand the authoring standard. - Copy
template/tofree-skills/<skill-name>/and write yourSKILL.md. - Run
python3 scripts/validate_skills.pyandpython3 scripts/generate_catalog.pyuntil both are clean. - Open a pull request.
📜 License
MIT — free to use, modify, and distribute. See LICENSE. Skills under
free-skills/anthropic/ retain their original upstream licenses (check the folder for a LICENSE.txt).
🙏 Acknowledgments
Built on Anthropic's open Agent Skills standard and Claude Code best
practices, with inspiration from the broader open-source skills community (including
obra/superpowers) and current, domain-specific research.
⭐ Star the repo · 📚 Catalog · 🐛 Issues · 💬 Discussions
Made for the agentic era.