# Can Claude Skills Be Used with AI Agents Other Than Claude? A Compatibility Guide > Discover if Claude Skills work with AI agents beyond Claude. Learn how this open standard ensures broad compatibility with tools like OpenAI Codex, Gemini CLI, and more. - Repository: [Composio/awesome-claude-skills](https://github.com/composiohq/awesome-claude-skills) - Tags: how-to-guide - Published: 2026-08-29 --- **Yes—Claude Skills function as an open standard that works across Claude Code, OpenAI Codex, Cursor, Gemini CLI, and other AI agents without requiring platform-specific modifications.** Claude Skills have evolved beyond the Anthropic ecosystem into a portable, language-agnostic format that any AI agent can adopt. As documented in the ComposioHQ/awesome-claude-skills repository, these skills rely on a standardized [`SKILL.md`](https://github.com/ComposioHQ/awesome-claude-skills/blob/main/SKILL.md) structure that multiple coding agents now recognize and execute, enabling you to reuse the same automation logic across different AI platforms. ## AI Agents That Support Claude Skills The repository's [`README.md`](https://github.com/ComposioHQ/awesome-claude-skills/blob/main/README.md) (lines 36 and 101) confirms that Claude Skills are now supported by a diverse range of agents. Because the specification is open, any agent implementing the Anthropic Skills spec can import and execute these skill definitions using the same underlying format. ### Claude Code and Claude.ai These remain the primary platforms for which the format was originally designed. Claude Code loads skills from the `~/.claude/skills/` directory or via the `--skill-dir` runtime flag. ### OpenAI Codex OpenAI's code-generation model behind GitHub Copilot can reference skills from prompts or `.codex` files. When a prompt mentions the skill's name, Codex fetches the [`SKILL.md`](https://github.com/ComposioHQ/awesome-claude-skills/blob/main/SKILL.md) contents and executes the defined workflow. ### Cursor The AI-first IDE treats [`SKILL.md`](https://github.com/ComposioHQ/awesome-claude-skills/blob/main/SKILL.md) files as executable task definitions. Cursor discovers these definitions in your workspace and runs them via the `cursor run` command, interpreting the skill's instructions just as it would native editor commands. ### Gemini CLI Google's command-line interface for Gemini models accepts a `--skill` argument pointing to a skill directory. The CLI reads the YAML frontmatter and body instructions to determine when and how to invoke the skill's capabilities. ### Antigravity and Windsurf These emerging agents load skills from configured `skills/` directories at startup, automatically discovering available capabilities without explicit installation commands. ## How the Open Standard Works A Claude Skill is fundamentally a directory containing a [`SKILL.md`](https://github.com/ComposioHQ/awesome-claude-skills/blob/main/SKILL.md) file. This file uses YAML front-matter to declare metadata (name, description, trigger conditions) followed by step-by-step workflow instructions. Because the format is language-agnostic, the agent parses the markdown structure rather than executing Claude-specific code. The typical skill structure looks like this: ``` pdf-processor/ ├─ SKILL.md # YAML frontmatter + workflow instructions ├─ scripts/ │ └─ process.py # Optional helper script used by the skill └─ references/ └─ docs.md # Optional documentation loaded on demand ``` According to the source code analysis, the skill's front-matter tells the agent *when* to invoke it, while the body holds the executable workflow. This separation allows any compliant agent to interpret the skill's intent regardless of its underlying model architecture. ## Loading Skills Across Different Agents While the skill format remains identical, each agent uses a slightly different command to load and execute skills: **Claude Code** uses the `--skill-dir` flag to point to the skill folder at runtime: ```bash claude --skill-dir ./pdf-processor ``` **Cursor** discovers skills in the workspace and executes them with the `run` subcommand: ```bash cursor run ./pdf-processor ``` **Gemini CLI** accepts the skill via a dedicated `--skill` argument: ```bash gemini --skill ./pdf-processor ``` **OpenAI Codex** references skills directly within prompts or `.codex` files using comments: ```bash /** Use skill: pdf-processor **/ ``` In all cases, the agent reads the same [`SKILL.md`](https://github.com/ComposioHQ/awesome-claude-skills/blob/main/SKILL.md) file, interprets the YAML frontmatter for context, and executes the markdown body as instructions. ## Key Files in the Repository Several files in the ComposioHQ/awesome-claude-skills repository demonstrate this cross-platform portability: - **[`README.md`](https://github.com/ComposioHQ/awesome-claude-skills/blob/main/README.md)** (lines 36-101): Documents the open-standard nature of Claude Skills and explicitly lists supported agents including Codex, Cursor, and Gemini CLI. - **[`mcp-builder/SKILL.md`](https://github.com/ComposioHQ/awesome-claude-skills/blob/main/mcp-builder/SKILL.md)**: Explains the Model Context Protocol (MCP) layer that enables skills to work with any agent supporting MCP, abstracting away platform-specific implementations. - **[`connect-apps-plugin/README.md`](https://github.com/ComposioHQ/awesome-claude-skills/blob/main/connect-apps-plugin/README.md)**: Shows how plugins enable real actions that can be adapted for non-Claude agents. - **Individual skill directories** (`composio-skills/*/SKILL.md`): Provide examples of reusable skill implementations that function identically across all supported platforms. ## Summary - Claude Skills operate as an **open standard** that multiple AI agents now support, not just Claude. - The same [`SKILL.md`](https://github.com/ComposioHQ/awesome-claude-skills/blob/main/SKILL.md) file works across **Claude Code, OpenAI Codex, Cursor, Gemini CLI, Antigravity, and Windsurf** without modification. - Skills rely on YAML frontmatter for metadata and markdown content for instructions, making them **language-agnostic** and **portable**. - Each agent loads skills through platform-specific commands (`--skill-dir`, `cursor run`, `--skill`), but interprets the same underlying format. - The repository's [`README.md`](https://github.com/ComposioHQ/awesome-claude-skills/blob/main/README.md) and [`mcp-builder/SKILL.md`](https://github.com/ComposioHQ/awesome-claude-skills/blob/main/mcp-builder/SKILL.md) document the protocol specifications that enable this cross-platform compatibility. ## Frequently Asked Questions ### Do I need to modify a Claude Skill to use it with Cursor or Gemini? No. Because Claude Skills follow an open specification, the same [`SKILL.md`](https://github.com/ComposioHQ/awesome-claude-skills/blob/main/SKILL.md) file and directory structure works across all supported agents. You can move a skill folder from Claude Code to Cursor or Gemini CLI without changing any files, as each agent parses the YAML frontmatter and markdown instructions identically. ### What file must be present for an AI agent to recognize a skill? Every skill requires a **[`SKILL.md`](https://github.com/ComposioHQ/awesome-claude-skills/blob/main/SKILL.md)** file at the root of the skill directory. This file must contain valid YAML frontmatter defining the skill's name and description, followed by the step-by-step instructions in markdown format. Agents discovered this specific filename to identify valid skill packages. ### Can I use Claude Skills with the OpenAI API directly? Yes, through OpenAI Codex. Codex can reference skills from prompts or `.codex` files by name. When you mention the skill in a comment or prompt, Codex retrieves the [`SKILL.md`](https://github.com/ComposioHQ/awesome-claude-skills/blob/main/SKILL.md) contents and executes the workflow. The repository confirms OpenAI Codex as a supported platform alongside Claude Code. ### Where should skills be stored for Claude Code specifically? Claude Code looks for skills in the **`~/.claude/skills/`** directory by default. You can also specify an alternate location at runtime using the `--skill-dir` flag followed by the path to your skill folder. This differs from Cursor and Gemini CLI, which typically read skills from the current workspace or require explicit path arguments.