yt-audience-insights
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content by retrieving and analyzing YouTube comments.
- Ingestion points: Untrusted text enters the context in
SKILL.mdunder Step 1, either dynamically fetched vialib.YouTubeClient().fetch_video_commentsor directly pasted by the user. - Boundary markers: Absent. The instructions do not define clear delimiters or add guidelines for the agent to ignore any potential prompt injections or instructions hidden within the analyzed comments.
- Capability inventory: The skill includes capabilities to read comments, extract themes, and automatically draft replies, which could be manipulated if a comment contains adversarial instructions.
- Sanitization: Absent. No structural validation or escaping of the comment text is performed prior to analysis.
Audit Metadata