skill-security-audit

To audit installed skills, run the security audit to validate frontmatter, scan body content, and review provenance across local and project paths, generating a comprehensive report with actionable remediation guidance.

Frontmatter validation verifies that SKILL.md files declare safe and precise capabilities, checking for misconfigured fields and ensuring skills do not contain unsafe instructions that could compromise agent behavior.

Detect prompt injection by auditing body content of installed skills to identify unsafe instructions, validating provenance, and reviewing Git history to produce a report documenting malicious or untrusted entries.

Yes, the security audit checks provenance across local and project skill repositories to identify sources, last modified dates, and Git history, supporting secure onboarding of new skills.

The best way to secure skill onboarding is running a comprehensive audit that validates frontmatter, checks body content for unsafe instructions, and reviews Git history to generate remediation guidance.

Unexpected agent behavior can stem from malicious or misconfigured SKILL.md files; auditing the skill library detects untrusted entries by validating frontmatter and checking body content for unsafe instructions.