This repository provides a centralized collection of reusable skills for coding agents within the OpenClaw ecosystem, addressing the problem of duplicated workflow logic across projects. It targets developers building or integrating with OpenClaw's agent framework, offering a significant benefit by enabling "write once, reuse everywhere" for common tasks like code review and remote validation, thereby reducing redundant effort and enhancing consistency.

How It Works

The project organizes reusable agent workflows into modular "skills," each residing in its own directory (skills/<name>/) containing a SKILL.md definition and associated helper scripts. An install-skills script facilitates local installation via symlinks (ideal for development) or file copies (for portable environments). It also supports direct integration with agent platforms like Codex and Claude by allowing the repository's skills directory to be symlinked into their respective skill paths, simplifying adoption and ensuring consistent access to shared functionalities.

Quick Start & Requirements

Clone the repository:

git clone https://github.com/openclaw/agent-skills.git
cd agent-skills

Install skills using scripts/install-skills. Options include listing (--list), dry-run (--dry-run), specifying a target directory (--target), using copies instead of symlinks (--mode copy), or forcing overwrites (--force).

For Codex integration, symlink the repo into ~/.codex/skills/agent-skills. For Claude Code integration, symlink into ~/.claude/skills/agent-skills or add individual skill symlinks within an existing shared skills folder.

No specific non-default prerequisites beyond standard scripting environments (bash, Python, Node.js) are detailed.

Highlighted Details

• Core Skills: Includes agent-transcript (local, redacted transcript provenance), autoreview (structured code review/closeout), crabbox (remote validation), handoff (task delegation), and session-viewer (local HTML session data viewer).
• Zero-Setup Repos: Important repositories may vendor a generated snapshot of skills under .agents/skills/<name> to provide immediate functionality for contributors without prior setup. These snapshots are distribution artifacts, not the source of truth.
• Validation: The scripts/validate-skills script enforces SKILL.md structure and checks script syntax across multiple languages (Ruby, Bash, Python, Node.js).
• Development Workflow: Symlinks are recommended for local development, ensuring changes in the checked-out repository are immediately reflected in the agent environment.

Maintenance & Community

The provided README does not detail specific contributors, sponsorships, or community channels (e.g., Discord, Slack).

Licensing & Compatibility

The project is licensed under the MIT license. This permissive license is generally compatible with commercial use and linking within closed-source projects.

Limitations & Caveats

Output from the session-viewer skill may contain sensitive conversation data and should be treated as local/private unless explicitly redacted. Vendored snapshots in downstream repositories are distribution artifacts; edits must be made in this canonical repository first, then synced. Project-specific skills should remain within their respective repositories.