How to Audit ClawHub Skills Before Installing
Quick Answer: Before installing any ClawHub skill, check the publisher's GitHub account age (must be 7+ days), read the skill.md file line by line for hidden install commands, verify star count and downloads, and run openclaw security audit --deep after installation. Snyk found 36% of ClawHub skills contain prompt injection risks.
With 10,700+ community skills on ClawHub, vetting is essential. The ClawHavoc attack identified 1,184 malicious packages designed to steal API keys and exfiltrate sensitive data.
Why Is ClawHub Skill Auditing Critical?
ClawHub is an open marketplace. Anyone can publish a skill, and malicious actors exploit this.
36%
of skills contain prompt injection risks
Snyk security research, 2026
1,184
malicious packages found in ClawHavoc
Coordinated supply chain attack
10,700+
total community skills on ClawHub
Growing daily — vetting cannot be automated fully
What Is the 5-Step ClawHub Skill Audit Checklist?
Follow these steps every time you install a new skill. It takes 5-10 minutes and can save you from a serious security breach.
Step 1: Check Publisher Account Age
Verify the publisher's GitHub account is at least 7 days old. New accounts are a major red flag — the ClawHavoc attack used freshly created accounts to publish malicious skills.
- Visit the publisher's GitHub profile
- Check account creation date (must be 7+ days old)
- Look for other repositories and contribution history
- Verify the account has not been recently renamed
Step 2: Read the skill.md File Line by Line
The skill.md file defines what the skill can do and what permissions it requests. Read every line — malicious skills hide install commands, network requests, and data exfiltration instructions in seemingly innocent descriptions.
- Open the skill.md file in the skill's repository
- Search for any curl, wget, or fetch commands
- Look for base64-encoded strings (used to hide payloads)
- Check for instructions that override OpenClaw's safety settings
Step 3: Verify Star Count and Downloads
Popular skills with hundreds or thousands of stars are generally safer — but not immune. Compare the star count to download count. A skill with many downloads but few stars may be artificially inflated.
- Check GitHub stars (higher is generally better)
- Compare stars to ClawHub download count
- Look for recent star spikes (may indicate manipulation)
- Read recent issues and pull requests for security reports
Step 4: Scan for Hidden Install Scripts
Check the skill's package.json or install scripts for post-install hooks that execute code during installation. This is the most common attack vector — code runs before you even use the skill.
- Check package.json for 'postinstall' or 'preinstall' scripts
- Look for shell scripts that run during installation
- Search for obfuscated JavaScript or encoded payloads
- Verify all dependencies are from known, trusted sources
Step 5: Run the Security Audit Command
After installation, always run OpenClaw's built-in security audit. This scans for known vulnerabilities, suspicious patterns, and unauthorized network access attempts.
- Run: openclaw security audit --deep
- Review each finding and its severity rating
- Address any CRITICAL or HIGH findings before using the skill
- Re-run after skill updates to catch new vulnerabilities
What Are the Red Flags to Watch For?
If you spot any of these warning signs, do not install the skill. Report it to ClawHub instead.
The ClawHavoc Attack: What Happened
In early 2026, security researchers discovered a coordinated supply chain attack on ClawHub. Dubbed "ClawHavoc," the attack involved 1,184 malicious packages uploaded by a network of fake accounts. Here is what they did:
- Typosquatting: Named skills similarly to popular ones (e.g., 'openclaw-slak' instead of 'openclaw-slack')
- Hidden install scripts: Post-install hooks that silently exfiltrated environment variables including API keys
- Prompt injection: skill.md files containing instructions that overrode OpenClaw's safety guardrails
- Data exfiltration: Captured and sent API keys, file contents, and system information to attacker-controlled servers
- Persistence: Some skills installed background processes that survived OpenClaw restarts
Learn more about protecting yourself: ClawHub Malicious Skills Report
How Do You Report a Suspicious Skill?
If you find a skill that looks malicious, report it immediately to protect the community.
Flag on ClawHub
Click the 'Report' button on the skill's ClawHub page. Include specific details about what you found suspicious.
Open a GitHub Issue
File an issue on the OpenClaw GitHub repository with the skill name, publisher, and evidence of malicious behavior.
Run and Share Audit Results
Run 'openclaw security audit --deep' and include the output in your report. This helps the security team triage quickly.
Notify the Community
Post in OpenClaw's Discord or community forums to warn other users while the report is being reviewed.
Frequently Asked Questions
Stop Wasting 40-60% of Your AI Budget
Download the free '6 Token Drains' guide — identify the hidden patterns burning through your tokens and get copy-paste fixes for each one.
Read the Free GuideYour Competitors Are Already Automating. Are You?
Every week we send one automation that saves 10+ hours of manual work — the same playbooks our clients use to run their businesses on autopilot. Miss a week, miss the edge.
Get the Automation Playbook (Free)
One deploy-ready automation every week. Same strategies our clients pay thousands for. 400+ business owners already inside.
Need it done for you?
Book a Free Strategy Call See what we've built for real businesses →