skill-security-scanner

You can scan AI agent skills for security vulnerabilities by running this tool against the skill directory to perform static analysis, threat intelligence checks, and LLM-based semantic analysis for malicious code.

LLM semantic analysis detects hidden threats by understanding code intent rather than just pattern matching. It evaluates the logic and behavior of AI agent skills to uncover disguised exploits and malicious activities.

Yes, you can integrate security scanning into your deployment pipeline using the provided REST API. The skill also offers command-line and web interface options to automate vulnerability checks during deployment.

Yes, the scanner provides standardized risk quantification using CVSS 3.1 scoring. This allows you to systematically evaluate and prioritize the 18 types of security risks and vulnerabilities detected in AI agent skills.

Yes, you need Python 3.9 or higher along with required libraries like requests, pyyaml, flask, and LLM client packages to execute the security scanning, static analysis, and deep semantic evaluation processes.

Security scanning identifies 18 types of security risks, including hardcoded secrets, malicious code, and potential exploits. It uses static analysis and threat intelligence to secure the entire AI agent skill lifecycle.