claw-skill-guard

What This Skill Does

claw-skill-guard is an essential security utility for every OpenClaw user. It serves as a static analysis engine designed to inspect the source code and installation manifests of agent skills before they are executed in your environment. By scanning for malicious patterns such as remote code execution via pipes, obfuscated base64 payloads, and unauthorized sudo calls, this tool acts as your first line of defense against supply chain attacks. It cross-references installation scripts against known-safe patterns and flags dangerous behaviors, providing a clear risk assessment (CRITICAL, HIGH, MEDIUM, LOW) so you can make informed decisions before running untrusted code.

Installation

To install the security scanner, add the following to your AGENTS.md file:

• Install Command: clawhub install openclaw/skills/skills/vincentchan/claw-skill-guard
• Source Repo: openclaw/skills
• Author: vincentchan

Once installed, ensure that your local OpenClaw directory is correctly mapped to the scripts/claw-skill-guard/ path so you can run the scanner from your terminal.

Use Cases

• Vetting Third-Party Plugins: Before adding a new productivity skill from an untrusted ClawHub author, run this scanner to ensure the plugin doesn't contain hidden backdoors or unauthorized network calls.
• Security Auditing: Regularly audit your local skills/ directory to ensure existing plugins haven't been modified or compromised through local file manipulation.
• Team Policy Enforcement: Integrate this into your development workflow to enforce a 'no install' policy for any skill that receives a CRITICAL or HIGH risk rating from the scanner.

Example Prompts

1. "Check the skill at https://clawhub.com/dev/auto-deploy-tool for any malicious install commands or suspicious network requests."
2. "Scan my local ./skills directory for any plugins that might be accessing .env files or attempting to escalate privileges."
3. "Is the 'twitter-analytics' skill safe to install? Run the claw-skill-guard scanner and summarize the findings."

Tips & Limitations

• Static Analysis Only: The scanner looks at code patterns but does not execute the code in a sandbox. It cannot catch highly sophisticated 'zero-day' exploits that do not follow known malicious patterns.
• False Positives: Sometimes, legitimate build tools look like malicious ones (e.g., a build tool using curl to fetch a dependency). Always manually review lines flagged as CRITICAL.
• Keep It Updated: Since malware techniques evolve, ensure you pull the latest version of the scanner frequently to get updated regex patterns.