OpenClaw is an open-source AI coding agent designed to streamline your development workflow by leveraging reusable AI skills defined in SKILL.md files. Understanding where OpenClaw stores these skill files, how to share them with Claude Code, and how to verify that your agent has loaded them correctly is crucial for maximizing your productivity. This guide dives deep into the OpenClaw skills directory path, environment overrides, cross-agent compatibility, and practical management tips to help you get the most out of your AI coding assistant.

Default OpenClaw Skills Paths

OpenClaw organizes its skills in a straightforward directory structure, supporting both personal and project-scoped skill sets. By default, skills are stored in two main locations depending on their scope: | Scope | Path | |---|---| | Personal (all projects) | `~/.openclaw/skills/` | | Project-scoped | `.openclaw/skills/` (inside the repository root) | The personal skills directory is intended for skills that you want available across all projects and sessions. This is your global skill library. On the other hand, the project-scoped skills directory lives inside each project repository and is ideal for project-specific or experimental skills that shouldn’t interfere with other projects. Depending on your operating system, the `~` expands differently: - macOS: `/Users/your-username/.openclaw/skills/` - Linux: `/home/your-username/.openclaw/skills/` - Windows: `C:\Users\your-username\.openclaw\skills\` This consistent path structure makes it easy to locate and manage your skill files regardless of your development environment.

Understanding the SKILL.md Format

OpenClaw uses SKILL.md files to define AI skills, a format that is fully compatible with Claude Code. Each skill directory contains a `SKILL.md` file that describes the skill’s purpose, triggers, prompts, and other metadata in a standardized way. This uniformity allows skills to be portable between different AI coding agents without modification. A typical `SKILL.md` file includes sections like: - Skill Name and Description - Activation phrases or triggers - The AI prompt template - Example usage This structure not only documents the skill clearly but also enables the agent to parse and execute the skill reliably.

Sharing Skills Between Claude Code and OpenClaw

Because OpenClaw and Claude Code share the same skill directory layout and SKILL.md format, you can seamlessly share skills between the two agents. If you already have a skills library for Claude Code, you don't need to rewrite or restructure anything to use those skills in OpenClaw.

Practical Sharing Techniques

- Symlinking: Create a symbolic link from the OpenClaw skills directory to your Claude Code skills folder. For example, on macOS/Linux, run `ln -s /path/to/claude/skills ~/.openclaw/skills` to link the directories. This keeps the skills in one place and any updates are instantly reflected. - Copying: If you prefer separate skill sets or want to freeze a snapshot of your Claude skills, simply copy the entire skills folder into OpenClaw’s skills directory. - Version Control: Since project-scoped skills live inside the repo, adding `.openclaw/skills/` to your version control system like Git ensures your team shares the exact same skill set.

Overriding Default Paths with Environment Variables

OpenClaw allows you to customize where it looks for skills by setting environment variables, which is useful if you want to maintain skills on external drives, cloud mounts, or custom directories. You can override the personal skills directory path by setting the `OPENCLAW_SKILLS_DIR` environment variable. For example: - On macOS/Linux (bash/zsh): `export OPENCLAW_SKILLS_DIR=/custom/path/to/skills` - On Windows (PowerShell): `setx OPENCLAW_SKILLS_DIR C:\custom\path\to\skills` Once set, OpenClaw will load your personal skills from this custom location instead of the default `~/.openclaw/skills/`. This flexibility is also useful in CI/CD environments or containers where the home directory might not be persistent or accessible. You can script your agent setup to point to a shared network location or mounted volume containing your skill set.

Verifying Skills Loaded Correctly

After placing your SKILL.md files in the appropriate directory, it’s important to verify that OpenClaw has successfully loaded them. Here’s a step-by-step approach: 1. Start OpenClaw: Launch your OpenClaw agent instance. 2. List Loaded Skills: Use the built-in command or interface option (depending on your setup) to list all active skills. This might be a CLI command like `openclaw skills list` or a UI panel that shows loaded skills. 3. Check Skill Metadata: Confirm that the skill names and descriptions match what you expect from the SKILL.md files. 4. Test Skill Activation: Trigger a skill by using one of its activation phrases or commands to ensure it runs as expected. If a skill does not appear, double-check the directory path, file permissions, and SKILL.md formatting. Logs or debug output from OpenClaw can provide clues about loading errors.

Real-World Tips for Managing Your Skills

- Organize Skills by Category: Inside your skills directory, use subfolders to group related skills (e.g., `git-tools`, `docker-helpers`, `python-utils`). This keeps your skill library tidy and easier to navigate. - Use Version Control: Especially for project-scoped skills, maintaining the `.openclaw/skills/` directory in Git ensures reproducibility and team collaboration. - Regularly Update Shared Skills: If you share skills between Claude Code and OpenClaw, keep the linked or copied directories in sync to avoid inconsistencies. - Document Custom Skills Thoroughly: Well-written SKILL.md files with clear examples make it easier for you and your team to understand and maintain skills over time. - Leverage Environment Overrides for Multi-User Machines: On shared development environments, setting `OPENCLAW_SKILLS_DIR` per user can prevent conflicts and ensure personalized skill sets.

Conclusion

Mastering the OpenClaw skills directory path and understanding how to share and verify your AI coding skills will unlock the full potential of your AI assistant. Whether you’re building a personal library of utilities or collaborating with a team using project-scoped skills, following these best practices ensures your skills are organized, accessible, and interoperable between OpenClaw and Claude Code. By combining directory management, environment configuration, and diligent verification, you'll create a robust AI skill ecosystem tailored to your development needs.