개요
Agent Skills는 절차적 지식과 컨텍스트를 휴대용 폴더로 패키징하여 에이전트가 필요에 따라 로드할 수 있는 오픈 표준입니다. 이 표준은 메타데이터와 지침을 포함하는 간단한 SKILL.md 형식을 정의하며, 호환되는 AI 에이전트 간에 도메인 전문 지식과 반복 가능한 워크플로우를 구현할 수 있게 합니다.
README 미리보기
# Agent Skills\n\n[](https://discord.gg/MKPE9g8aUy)\n\nA standardized way to give AI agents new capabilities and expertise.\n\n## What are Agent Skills?\n\nAgent Skills are a lightweight, open format for extending AI agent capabilities with specialized knowledge and workflows.\n\nAt its core, a skill is a folder containing a `SKILL.md` file. This file includes metadata (`name` and `description`, at minimum) and instructions that tell an agent how to perform a specific task. Skills can also bundle scripts, reference materials, templates, and other resources.\n\n```\nmy-skill/\n├── SKILL.md # Required: metadata + instructions\n├── scripts/ # Optional: executable code\n├── references/ # Optional: documentation\n├── assets/ # Optional: templates, resources\n└── ... # Any additional files or directories\n```\n\n## Why Agent Skills?\n\nAgents are increasingly capable, but often don't have the context they need to do real work reliably. Skills solve this by packaging procedural knowledge and company-, team-, and user-specific context into portable, version-controlled folders that agents load on demand. This gives agents:\n\n- **Domain expertise**: Capture specialized knowledge — from legal review processes to data analysis pipelines to presentation formatting — as reusable instructions and resources.\n- **Repeatable workflows**: Turn multi-step tasks into consistent, auditable procedures.\n- **Cross-product reuse**: Build a skill once and use it across any skills-compatible agent.\n\n## How do Agent Skills work?\n\nAgents load skills through **progressive disclosure**, in three stages:\n\n1. **Discovery**: At startup, agents load only the name and description of each available skill, just enough to know when it might be relevant.\n\n2. **Activation**: When a task matches a skill's description, the agent reads the full `SKILL.md` instructions into context.\n\n3. **Execution**: The agent follows the instructions, optionally exe