OpenClaw Skills
    openclaw
    install
    skills

    How to Install Skills in OpenClaw

    Use expert AI agent skills through Agensi MCP. Connect your agent, choose a skill and load its latest instructions on demand.

    April 23, 20265 min read
    Share:

    A skill is a set of instructions packaged as a SKILL.md file that an AI agent reads to learn a new workflow. OpenClaw uses the same SKILL.md standard as Claude Code, so installation follows the same pattern with slightly different paths.

    Quick Answer: Agensi skills load through your MCP-connected OpenClaw agent when you use them. The local paths and GitHub commands below are for your own or externally licensed skills, not Agensi purchases.

    Commands and paths at a glance

    WhatValue
    Personal skills directory~/.openclaw/skills/
    Project skills directory.openclaw/skills/ (repo root)
    Required file per skill<skill-name>/SKILL.md
    Install from zipunzip skill.zip -d ~/.openclaw/skills/
    Use from AgensiConnect through MCP and call get_skill on each use
    Install from GitHubgit clone https://github.com/user/skill.git ~/.openclaw/skills/skill
    Reuse a Claude Code skillcp -r ~/.claude/skills/code-reviewer ~/.openclaw/skills/
    List installed skillsls ~/.openclaw/skills/
    Apply changesstart a new OpenClaw session

    Every command above assumes OpenClaw is installed (openclaw --version returns a version). OpenClaw also ships with over 100 built-in AgentSkills covering file management, shell commands, web browsing, and API calls, so your installed skills sit alongside those.

    What do I need before installing skills?

    Make sure OpenClaw is installed and working. Open your terminal and verify:

    openclaw --version
    

    If you get a version number, you're ready. You also need a skills directory. OpenClaw creates ~/.openclaw/skills/ automatically on first run, but you can create it manually:

    mkdir -p ~/.openclaw/skills
    

    How do I use Agensi skills?

    Connect your AI agent to Agensi through MCP, then ask it to use the skill on your task. Approve any required purchase. Your agent loads the latest approved version through get_skill on each use. Never write any part of the skill to disk; your licence does not allow you to save, copy or export it.

    How do I install skills from GitHub?

    Search GitHub for SKILL.md files or browse repositories that publish skills:

    # Clone the repo
    git clone https://github.com/username/my-skill.git
    
    # Copy the skill folder to your skills directory
    cp -r my-skill ~/.openclaw/skills/
    

    Make sure the SKILL.md file is at ~/.openclaw/skills/my-skill/SKILL.md, not nested deeper. A common mistake is ending up with ~/.openclaw/skills/my-skill/my-skill/SKILL.md, that won't work.

    Always read the SKILL.md before installing GitHub-sourced skills. Unlike Agensi marketplace skills, GitHub skills aren't security-scanned.

    How do I use Agensi skills?

    Connect your AI agent to Agensi through MCP, then ask it to use the skill on your task. Approve any required purchase. Your agent loads the latest approved version through get_skill on each use. Never write any part of the skill to disk; your licence does not allow you to save, copy or export it.

    How do I share skills with my team?

    Put skills in .openclaw/skills/ at the root of your project repository and commit them to git. Everyone who clones the repo gets the skills automatically.

    your-project/
    ├── .openclaw/
    │   └── skills/
    │       ├── code-reviewer/
    │       │   └── SKILL.md
    │       └── test-generator/
    │           └── SKILL.md
    ├── src/
    └── package.json
    

    Project-level skills override personal skills if they share the same name. This lets teams enforce specific workflows without affecting developers' personal skill setups.

    What are the most common installation mistakes?

    Double-nested folders. After unzipping, check that the path is ~/.openclaw/skills/skill-name/SKILL.md, not ~/.openclaw/skills/skill-name/skill-name/SKILL.md.

    Missing SKILL.md file. A folder without SKILL.md inside it is ignored. The file must be named exactly SKILL.md (case-sensitive on Linux/Mac).

    Wrong directory. OpenClaw uses ~/.openclaw/skills/, not ~/.claude/skills/. If you're switching between agents, double-check the path. Both agents use the same SKILL.md format, but they read from different directories.

    Not starting a new session. Skills load at session startup. If you install a skill mid-session, it won't be available until you start a new one.

    Can I reuse my whole Claude Code skills folder?

    Yes. Instead of copying skills one by one you can symlink the entire directory:

    # Copy a specific skill
    cp -r ~/.claude/skills/code-reviewer ~/.openclaw/skills/code-reviewer
    
    # Or symlink every Claude skill into OpenClaw
    ln -s ~/.claude/skills/* ~/.openclaw/skills/
    

    Most skills work across both agents without any change, because the SKILL.md format is identical.

    Which skills need adjustment for OpenClaw?

    The format is the same, but a few behaviours differ:

    • Skills using Claude Code's context: fork setting to run as subagents may behave differently, since OpenClaw handles concurrent tasks through its own gateway system.
    • Skills that call Claude-specific slash commands (like /simplify or /batch) won't work, because those are built into Claude Code.
    • Skills that stick to plain markdown instructions, shell commands, and file operations behave identically on both platforms.

    How do I create a skill for OpenClaw?

    mkdir -p ~/.openclaw/skills/my-skill
    

    Then create SKILL.md inside it. The frontmatter is identical to Claude Code:

    ---
    name: my-skill
    description: Description of what this skill does and when to use it.
    ---
    
    # My Skill
    
    Instructions here.
    

    What should I check before installing an untrusted skill?

    OpenClaw is open source, so anyone can publish skills, and community repositories have had security incidents, including malicious skills performing data exfiltration and prompt injection. Before installing from an untrusted source:

    • Read the whole SKILL.md.
    • Check every file in the skill's scripts/ directory.
    • Be sceptical of skills that make network requests or read credentials and sensitive files.

    Marketplaces like Agensi run an automated security scan on every submission, which is why a marketplace copy of a skill is safer than a random repo. See how the Agensi security scan works.

    Can I use the same skills in Claude Code?

    Yes. SKILL.md is a cross-agent standard. The same skill file works in OpenClaw, Claude Code, Codex CLI, Cursor, and Gemini CLI. Just copy it to the appropriate skills directory for each agent:

    • OpenClaw: ~/.openclaw/skills/
    • Claude Code: ~/.claude/skills/
    • Codex CLI: ~/.codex/skills/
    • Cursor: .cursor/skills/ (project-level only)

    For a full path reference, read Where Are Claude Skills Stored?.


    Find and install security-scanned skills for OpenClaw, Claude Code, and more on Agensi.

    Keep reading

    Frequently Asked Questions

    Connect your AI agent to Agensi through MCP. Ask it to use the skill and approve any required purchase. Instructions load on each use. Never write any part of the skill to disk.

    Ensure OpenClaw is installed and working by checking its version. You also need a skills directory, which OpenClaw creates automatically as ~/.openclaw/skills/ on its first run, or you can create it manually.

    Connect your AI agent to Agensi through MCP. Ask it to use the skill and approve any required purchase. Instructions load on each use. Never write any part of the skill to disk.

    Clone the GitHub repository, then copy the skill's folder into your ~/.openclaw/skills/ directory. Ensure the SKILL.md file is not nested deeper than one folder level within the `skills` directory.

    Place skill folders with their SKILL.md files in a `.openclaw/skills/` directory at the root of your project repository and commit them to git. This makes the skills available to anyone who clones the repo.

    Common mistakes include double-nested folders, a missing or incorrectly named SKILL.md file, using the wrong skills directory path, or not starting a new OpenClaw session after installation.

    Yes, SKILL.md is a cross-agent standard, so the same skill file works in OpenClaw, Claude Code, Codex CLI, and Gemini CLI. Copy it to the respective agent's skills directory, such as ~/.claude/skills/ for Claude Code.

    Connect your AI agent to Agensi through MCP. Ask it to use the skill and approve any required purchase. Instructions load on each use. Never write any part of the skill to disk.