# AI Agent Skills AI Agent Skills is an open standard specification for giving [[AI Agents]] reusable capabilities and domain knowledge. Originally developed by [[Anthropic]] for [[Claude Code]], the format was released as an open standard on December 18, 2025, and has since been adopted across the industry by competing platforms, following a similar trajectory to [[Model Context Protocol (MCP)]]. The specification is maintained at https://agentskills.io and on GitHub at https://github.com/agentskills/agentskills. ## Core Concept A skill is a directory containing a `SKILL.md` file with YAML frontmatter and Markdown instructions. The frontmatter describes what the skill does and when to use it. The Markdown body contains the instructions an agent follows when the skill is activated. Skills can also bundle scripts, reference documentation, and static assets. The key idea is **progressive disclosure**: agents load only the skill's name and description at startup (~100 tokens), then load full instructions only when the skill is activated, and pull in supporting files only when needed. This keeps context usage efficient. ## Specification The `SKILL.md` frontmatter requires two fields: | Field | Required | Description | |-------|----------|-------------| | `name` | Yes | Lowercase letters, numbers, hyphens. Max 64 characters. Must match directory name | | `description` | Yes | What the skill does and when to use it. Max 1024 characters | | `license` | No | License applied to the skill | | `compatibility` | No | Environment requirements | | `metadata` | No | Arbitrary key-value pairs for platform-specific extensions | | `allowed-tools` | No | Pre-approved tools the skill may use (experimental) | Directory structure: ``` skill-name/ ├── SKILL.md # Required ├── scripts/ # Executable code ├── references/ # Additional documentation └── assets/ # Templates, images, data files ``` ## Industry Adoption Within two months of the open standard's release, [[OpenAI]] added skills support to both ChatGPT and their Codex CLI tool, adopting a structurally identical architecture with the same file naming conventions and metadata format. OpenAI extends the standard with an optional `agents/openai.yaml` file for UI metadata. As of early 2026, the following platforms support Agent Skills: **AI coding tools**: [[Claude Code]], [[Cursor.com]], [[GitHub]] Copilot (via VS Code), [[OpenAI Codex]], Roo Code, [[Windsurf]], Amp, [[Goose]] (by Block), [[Gemini CLI]], Mistral AI Vibe, Trae, Factory, OpenCode, Autohand, Command Code, Mux, Firebender, VT Code, Piebald **AI platforms and frameworks**: Claude.ai, Letta, Databricks, Spring AI, Qodo, Agentman, Ona The breadth of adoption across competing vendors (Anthropic, OpenAI, Google, Mistral, Microsoft/GitHub) makes Agent Skills one of the few genuinely cross-platform standards in the AI agent ecosystem. ## Comparison with MCP Agent Skills and [[Model Context Protocol (MCP)]] are complementary: - **MCP** provides a standard for connecting agents to external tools and data sources (the "how") - **Agent Skills** provides a standard for packaging instructions and domain knowledge (the "what to do") Both standards originated at Anthropic and were opened to the broader ecosystem. Simon Willison has speculated that Agent Skills may eventually join the [[Agentic AI Foundation (AAIF)]], which already hosts MCP. ## References - Specification: https://agentskills.io/specification - GitHub repository: https://github.com/agentskills/agentskills - Example skills: https://github.com/anthropics/skills - Reference library (validation): https://github.com/agentskills/agentskills/tree/main/skills-ref - OpenAI Codex skills: https://developers.openai.com/codex/skills - Simon Willison's analysis: https://simonwillison.net/2025/Dec/19/agent-skills/ ## Related - [[NVIDIA SkillSpector]] - [[Claude Code]] - [[Claude Code Skills]] - [[OpenAI Codex]] - [[Gemini CLI]] - [[AI Agents]] - [[Anthropic]] - [[OpenAI]] - [[Agentic AI Foundation (AAIF)]] - [[Model Context Protocol (MCP)]] - [[Large Language Models (LLMs)]] - [[Context7]] - [[AI Agents Web Browsing]] - [[Browser Use]] - [[Vercel Agent Browser]] - [[AI skills operationalize passive knowledge]] - [[My Published AI Skills]] - [[AI Agent Skills in Chrome]]