skill-vetter

To vet AI agent skills for malware, you need a standardized security review process that scans for malicious code patterns, evaluates permission scope, and classifies risk levels. This involves source verification, mandatory code review, and red flag detection to prevent credential theft and data exfiltration prior to installation.

During a code review of third-party agent skills, look for malicious patterns like unauthorized network calls, credential file access, and obfuscated code. Identifying these red flags helps prevent unauthorized data exfiltration and malicious code execution within your automated agent workflows.

Yes, a standardized security vetting protocol applies to skills sourced from GitHub, ClawdHub, or shared by other agents. It evaluates permission scope and detects malicious code patterns across all sources to ensure safe integration into automated agent workflows.

The best way to evaluate the permission scope of an untrusted AI agent skill is through a standardized vetting protocol that includes source verification, mandatory code review, and risk classification. This process identifies unauthorized access to credentials and sensitive data before installation.

You need to run a security risk assessment on task automation skills prior to installation whenever they come from untrusted sources like unknown GitHub repositories. This prevents malware infection, credential theft, and sensitive data exfiltration by classifying risk levels before integration.

Agent safety vetting prevents unauthorized network calls by scanning all files in shared task automation skills for suspicious behavior. It generates a formal security report that flags malicious code patterns and classifies risk levels for informed installation decisions.