comfyui-research
Warn
Audited by Snyk on Mar 1, 2026
Risk Level: MEDIUM
Full Analysis
MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).
- Third-party content exposure detected (high risk: 0.90). The skill explicitly fetches and ingests untrusted, user-generated content (YouTube video transcripts via youtube-video-analyst, GitHub release notes, HuggingFace model pages, and subreddit posts listed in references/watch-list.yaml) and uses that content to update reference files and drive research decisions, so third-party instructions could materially influence agent behavior.
Audit Metadata