OpenClaw Skills: Complete Guide to ClawHub, Installation, and Alternatives

OpenClaw Skills: Complete Guide to ClawHub, Installation, and Alternatives

Quick Answer: OpenClaw is an open-source AI agent framework with 68,000+ GitHub stars and 13,000+ community skills on ClawHub. Skills install via clawhub install <skill-slug> and live in ~/.openclaw/skills/. The same SKILL.md format works across OpenClaw, Claude Code, Cursor, and Codex CLI. For curated, security-scanned alternatives to ClawHub's unfiltered catalog, browse Agensi.

---

OpenClaw (originally called Clawdbot) is the largest open-source AI agent framework by GitHub stars. Created by PSPDFKit founder Peter Steinberger, it runs as a local Node.js service that connects AI models to your tools through chat platforms like Telegram, Discord, and WhatsApp.

The skill ecosystem around OpenClaw has grown fast. ClawHub, the community skill registry, now catalogs over 13,000 skills. That scale is both OpenClaw's strength and its biggest challenge for new users.

What is ClawHub?

ClawHub is OpenClaw's community skill marketplace. Anyone can publish a skill, and the registry indexes everything without formal review. As of May 2026, it tracks over 13,729 skills across categories like data science, productivity, DevOps, and finance/trading.

Skills install from ClawHub with a single command:

clawhub install <skill-slug>

The skill downloads to ~/.openclaw/skills/ and is available on the next session.

How to install OpenClaw skills

From ClawHub

# Install a skill
clawhub install capability-evolver

# List installed skills
ls ~/.openclaw/skills/

# Remove a skill
rm -rf ~/.openclaw/skills/capability-evolver

From Agensi (security-scanned)

mkdir -p ~/.openclaw/skills && curl -sL https://www.agensi.io/api/install/<slug> | tar xz -C ~/.openclaw/skills/

From GitHub

git clone https://github.com/author/skill-name.git
cp -r skill-name ~/.openclaw/skills/

Skill directory structure

OpenClaw mirrors Claude Code's directory layout:

~/.openclaw/skills/          # Personal skills (all projects)
.openclaw/skills/            # Project-scoped skills (per repo)

Each skill is a folder containing a SKILL.md file:

~/.openclaw/skills/
├── code-reviewer/
│   └── SKILL.md
├── test-generator/
│   └── SKILL.md
└── env-doctor/
    ├── SKILL.md
    └── scripts/
        └── check-deps.sh

The ClawHub quality problem

13,000+ skills sounds impressive until you start installing them. A security audit published in early 2026 found an average of 6.3 issues per skill across major registries including ClawHub. Snyk's ToxicSkills research found prompt injection patterns in 36% of skills tested.

The problem is structural: ClawHub has no review process. Anyone publishes anything. That means:

• Many skills are abandoned experiments or half-finished prototypes
• No security scanning before publication
• Quality ranges from production-grade to broken
• Duplicate and near-duplicate skills clutter search results

This doesn't make ClawHub unusable. It means you need to vet every skill before installing. Read the SKILL.md, check the author's other work, and look at install counts as a basic signal.

Top OpenClaw skill categories

Based on ClawHub data as of May 2026:

The category distribution is notably different from Claude Code's ecosystem, where DevOps leads. OpenClaw's strength in data science reflects its origins as a personal automation tool rather than a coding agent.

OpenClaw vs Claude Code skills

The SKILL.md format is identical across both platforms. The same file works in both agents without modification. The differences are in the runtime:

What's the same:

• SKILL.md format (YAML frontmatter + markdown body)
• Directory structure (~/.agent/skills/ for personal, .agent/skills/ for project)
• Skill activation based on description matching
• Support for scripts, references, and assets

Where they differ:

• OpenClaw supports multi-agent orchestration natively, running skills across connected chat platforms
• Claude Code has deeper IDE integration and subagent spawning via context: fork
• OpenClaw's skill discovery is more aggressive (loads more skills into context)
• Claude Code's allowed-tools permission system doesn't exist in OpenClaw

Portability: Copy a skill folder between ~/.claude/skills/ and ~/.openclaw/skills/ and it works in both. No file changes needed.

ClawHub vs Agensi: choosing a skill source

Use ClawHub when you want the broadest selection and are comfortable vetting skills yourself. Good for discovery and experimentation.

Use Agensi when you want skills that have been security-reviewed, or when you're working across multiple agents and want portable skills with a consistent quality bar.

In practice, experienced OpenClaw users pull from both: ClawHub for quick experiments and niche use cases, Agensi for skills they'll rely on in production workflows.

Getting started with OpenClaw skills

If you're new to OpenClaw, start with these high-value skill categories:

1. Productivity automation — email triage, meeting summaries, task prioritization
2. Data processing — CSV cleaning, notebook conversion, data analysis
3. DevOps — server health checks, deployment automation, log analysis
4. Communication — message drafting, translation, content repurposing

Install 2-3 skills that match your daily workflow, test them for a week, then expand. Don't install 50 skills at once — context window bloat reduces quality across the board.

For curated, security-scanned skills that work with OpenClaw and every other SKILL.md-compatible agent, browse the Agensi marketplace.

---

OpenClaw skills use the same SKILL.md format as Claude Code, Cursor, and Codex CLI. Browse cross-compatible skills at agensi.io/skills.