skill-scanner

To scan agent skills for prompt injection, run the scanner against the target path. It analyzes SKILL.md files, scripts, and references using static analysis and pattern matching to detect malicious code and excessive permissions.

When auditing a community-developed agent skill, check for prompt injection, malicious scripts, secret exposure, excessive permissions, and supply chain risks. Analyzing SKILL.md and associated scripts validates frontmatter and assesses URL trustworthiness to ensure safety guidelines are met.

Yes, the skill scanner validates YAML frontmatter in SKILL.md files and works alongside Python environments. It requires the pyyaml dependency to parse configuration data and check for dangerous code patterns within skill scripts.

To check agent skill scripts for excessive permissions and secret exposure, use static analysis to scan the codebase. The scanner identifies dangerous code patterns, validates frontmatter, and flags potential secret exposure before deployment.

The best way to assess URL trustworthiness in agent skill references is to use an automated scanner that evaluates links during static analysis. This process assesses supply chain risks by validating external URLs against known threat vectors.