agents-libagents-lib
AgentsMCP ServersWorkflowsTutorialsComparisonsTemplatesNewsAbout
AG

Agent Skills Specification

⭐ 22,900•Markdown•Specification

Community-driven open standard for agent skills across AI agent platforms.

GitHubWebsite →
SpecificationAgent SkillsInteroperabilityOpen StandardCommunity

Overview

The Agent Skills Specification defines the format for agent skills that can be shared across platforms like Claude Code, Gemini CLI, and Cursor. With nearly 23,000 GitHub stars, it has become the reference implementation enabling interoperability in the agent skills ecosystem.

Features

  • ✓Open standard for skill format
  • ✓Cross-platform compatibility
  • ✓Discovery and validation protocol
  • ✓Composable skill architecture
  • ✓Semantic versioning support
  • ✓Rich metadata for discovery

Installation

Follow specification to create compatible skills

Pros

  • +Enables platform interoperability
  • +23,000+ stars indicating strong adoption
  • +Clear specification for developers
  • +Validation framework for quality
  • +Community-maintained and evolving

Cons

  • −Requires agent platform support
  • −Specification still evolving
  • −Quality varies across skills
  • −Uneven platform adoption

Alternatives

Claude Code skills →Gemini skills →

Documentation

Agent Skills Specification

Overview

The Agent Skills Specification is a community-driven standard for defining, sharing, and consuming agent skills across different AI agent platforms. With nearly 23,000 GitHub stars, it has become the reference implementation for the Agent Skills open standard, enabling interoperability between agent platforms like Claude Code, Gemini CLI, and other coding agents.

The specification defines a structured format for skills that can be discovered, validated, and executed by any compliant agent platform. This standardization enables a thriving ecosystem where skills developed for one platform can be used across others.

Features

  • Open standard specification: Defines the format and behavior of agent skills
  • Cross-platform compatibility: Skills work across Claude Code, Gemini CLI, Cursor, and more
  • Discovery protocol: Skills can be discovered and installed automatically
  • Validation framework: Ensure skills meet quality and security standards
  • Composable architecture: Skills can be combined and extended
  • Documentation standard: Consistent documentation format for all skills
  • Versioning support: Semantic versioning for skill compatibility
  • Metadata system: Rich metadata for skill discovery and filtering

Installation

# Clone the specification repository
git clone https://github.com/agentskills/agentskills
cd agentskills

# Review the specification and examples
# Follow the specification to create compatible skills

Quick Start

A basic Agent Skill follows the specification format:

# My Agent Skill

---
name: my-skill
description: Does something useful for agents
version: 1.0.0
triggers: ["trigger-phrase"]
metadata:
  author: "Your Name"
  license: "MIT"
---

## Instruction

This skill does X when triggered.

## Usage

Agent instructions here...

Core Concepts

Skill Definition

Each skill is a self-contained markdown file with YAML frontmatter metadata and instructions, following the Agent Skills specification.

Trigger System

Skills are activated by specific trigger patterns, allowing agents to invoke the right skill for the right task.

Platform Agnosticism

The specification is designed to be platform-agnostic, enabling skills to work across different agent implementations.

Advanced Features

Skill Composition

Combine multiple skills to create more complex capabilities through the specification's composition patterns.

Validation Pipeline

Automated validation ensures skills conform to the specification and meet quality standards.

Registry Integration

Skills can be registered in community registries for easy discovery and installation.

Examples

Basic Skill

# Code Review Skill

---
name: code-review
description: Reviews code for quality and best practices
version: 1.0.0
triggers: ["review this code", "code review"]
metadata:
  author: "Community"
  license: "MIT"
---

## Instruction

When triggered, review the provided code for:
1. Code quality and style
2. Security vulnerabilities
3. Performance issues
4. Best practice adherence

## Usage

Provide code context and this skill will analyze it.

Complex Multi-Trigger Skill

# Research Skill

---
name: research
description: Conducts research on technical topics
version: 1.1.0
triggers: ["research", "investigate", "find information about"]
metadata:
  author: "Community"
  license: "MIT"
  categories: ["research", "information-gathering"]
---

## Instruction

When triggered, conduct systematic research on the topic...

Pros

  • ✅ Open standard enables platform interoperability
  • ✅ 23,000+ stars indicating strong adoption
  • ✅ Enables vibrant skill ecosystem
  • ✅ Clear specification for skill developers
  • ✅ Validation framework ensures quality
  • ✅ Community-maintained and evolving

Cons

  • ✅ Requires agent platform support for full benefit
  • ✅ Specification still evolving as ecosystem grows
  • ✅ Quality varies across community-contributed skills
  • ✅ Adoption across platforms is uneven

When to Use

Use the Agent Skills Specification when developing skills for AI agents that should work across multiple platforms, contributing to the shared skill ecosystem, or building agent platforms that want to support the standard skill format.

Resources