skill-vetter

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.70). SKILL.md describes static analysis of a “target skill” by scanning its files (e.g., SKILL.md, index.js, package.json) after the user installs/unpacks that skill, which would be outsider-authored free text/code from the downloaded third-party skill entering the agent’s LLM context during vetting.