youtube-transcript
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data (YouTube transcripts), which constitutes a potential attack surface.
- Ingestion points: Untrusted data enters the context through
agentbody_youtube_transcriptandyoutube_web_v2_video_captions_result. - Boundary markers: The skill provides instructions to separate factual observations from interpretation, though no strict technical delimiters are defined.
- Capability inventory: The agent is restricted to read-only research via the SandBase gateway and lacks access to sensitive files or shell execution.
- Sanitization: There are no explicit sanitization or filtering steps defined for the ingested transcript content.
Audit Metadata