skill-security-auditor

You can audit AI agent skills by scanning code, prompts, and dependencies for risky patterns like prompt injection and data exfiltration. The auditor returns a PASS, WARN, or FAIL verdict with actionable remediation guidance to ensure safe installation.

Before installing AI skills, check for code execution, prompt injection, data exfiltration, credential access, and file system abuse. The auditor validates these patterns and provides a structured report with severity levels and suggested fixes.

Yes, the security auditor supports both local skill directories and remote repository URLs. You can scan remote sources directly and output the findings in JSON format for automation pipelines.

To generate a JSON security report, run the auditor on a local skill path or repository URL and enable JSON output. The report includes findings, severity ratings, and remediation guidance in a structured format for automation.

Yes, the auditor includes a strict mode option for vulnerability scanning. This enforces tighter validation on risky patterns in code and prompts, providing a more conservative PASS or FAIL verdict.

Run a pre-install security audit whenever integrating new AI agent skills that require system or network access. This prevents risky code execution, credential access, and data exfiltration patterns from compromising your environment.