Know when to use it.
Good metadata tells an agent when a skill is relevant, so expertise arrives at the right moment instead of bloating every conversation.
THE AGENT SKILLS LIBRARYOPEN INDEX
Give your agent
a better way to work.
Skills are reusable capabilities: a clear set of instructions, references, and tools that help an agent handle one kind of work well. Find one for the task in front of you, inspect where it came from, and install it when it earns your trust.
START WITH
01HOW TO THINK ABOUT SKILLS
Less setup.
More leverage.
A good skill is not a personality costume or a wall of vague advice. It is a portable operating guide for a repeatable job.
Follow a real method.
The useful part is concrete: steps, examples, constraints, scripts, and references that make the result more repeatable.
See where it came from.
Every card keeps provenance visible. Check the repository, license, install history, and audits before giving a skill access to your agent.
02THE LIBRARY
Find the missing capability.
Imported from the public skills ecosystem as metadata-first records. Search by what you need to do, not by what the repository happens to call it.
1,650 skillsSorted by install history
OFFICIAL01
find-skills
Discover and install agent skills by task, domain, source reputation, and install history.
Inspect source ↗OFFICIAL02
frontend-design
Build distinctive, production-ready interfaces with deliberate visual systems and strong interaction details.
Inspect source ↗WEB & BROWSER03
agent-browser
Navigate, click, fill forms, extract page data, and automate browser workflows from an agent.
Inspect source ↗ENGINEERING04
vercel-react-best-practices
Apply Vercel engineering guidance for fast, maintainable React and Next.js applications.
Inspect source ↗03THE AGENT CONNECTION
One search.
Every agent.
Connect the EveryAI MCP gateway and let an agent search, inspect, and read skills without leaving its workspace.
READ-ONLY BY DEFAULT
Give your agent a skills index it can actually use.
The gateway exposes task search, metadata, provenance, source registries, and explicit skill-file inspection. It never silently installs or runs third-party code.
search_skillsget_skillread_skilllist_skill_sources
LOCAL MCP CONFIG
{
"mcpServers": {
"everyai-skills": {
"command": "node",
"args": [
"/path/to/everyai/mcp/server.mjs"
]
}
}
}04WHERE SKILLS COME FROM
One index.
Many homes.
EveryAI brings the ecosystem into one readable starting point while keeping each publisher, repository, and installation path visible.
INDEXSkills.sh
Open ecosystem leaderboard and search.
OFFICIALAnthropic SkillsReference skills for Claude and the open standard.
OFFICIALOpenAI SkillsCurated and experimental skills for Codex.
OFFICIALVercel Agent SkillsWeb, React, Next.js, and shipping workflows.
OFFICIALMicrosoft Agent SkillsAzure and Microsoft Learn grounded expertise.
REGISTRYClawHubA public registry for OpenClaw skills and versions.
VERIFIEDNVIDIA SkillsSigned, evaluated CUDA-X and platform skills.
CURATEDAwesome Claude SkillsCommunity workflows and integrations.
05THE SHORT VERSION
Questions people
actually ask.
What is an AI agent skill?
A skill is a portable folder—usually centered on a SKILL.md file—that gives an agent focused instructions and optional scripts, references, and assets for a repeatable task.
How do I install one?
Open a card, inspect the source, then run its visible command with the skills CLI. For example: npx skills add owner/repo --skill skill-name.
Are community skills safe?
Not automatically. Read the source, check provenance and audits, understand requested permissions, and test unfamiliar skills in a constrained environment before granting access to sensitive systems.
Which agents support the format?
The open format is used across modern coding and agent tools, including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Cline, OpenCode, and others. Install paths vary by agent.
Read the Agent Skills specification for the format itself. EveryAI indexes metadata; the publisher remains the source of truth for the full skill.