skillguard

You can scan an AI agent skill for security vulnerabilities by running a three-layer analysis engine against the local directory, ClawHub skill, or arbitrary text content to detect credential theft, code injection, and prompt manipulation.

An AI skill audit can detect credential theft, code injection, prompt manipulation, data exfiltration, persistence mechanisms, and evasion techniques using a three-layer analysis engine that includes pattern matching and AST analysis.

Yes, you need Node.js installed in your environment to execute the security audit on local skill directories, as the vulnerability scanning engine requires the Node.js runtime to perform its three-layer analysis.

Yes, you can analyze arbitrary text content for prompt manipulation and data exfiltration threats, alongside scanning local directories and ClawHub skills, to receive a comprehensive security score and detailed vulnerability report.

The best way to secure an AI agent from malicious skills is to run a vulnerability scan before installation, utilizing multi-layered analysis to catch code execution, evasion techniques, and prompt attacks, yielding a detailed security report.

Yes, SkillGuard detects evasion techniques by employing AST analysis alongside pattern matching and advanced prompt injection detection to uncover hidden credential theft, code injection, and data exfiltration mechanisms in agent skills.