youtube-clip
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data fetched from YouTube.
- Ingestion points: Metadata (title, channel) and captions (SRT files) are ingested from YouTube videos to determine video structure.
- Boundary markers: There are no specific delimiters or protective instructions used when presenting caption content to the agent to prevent it from being misinterpreted as commands.
- Capability inventory: The agent can execute shell commands via yt-dlp, ffmpeg, and jq, and can write to the local file system.
- Sanitization: No sanitization is mentioned for captions or metadata before they are processed by the agent.
- [COMMAND_EXECUTION]: The skill relies on executing system-level binaries with arguments that may include external data.
- Evidence: Commands such as yt-dlp and ffmpeg are used with video URLs and metadata-derived values, creating a command execution surface that requires careful handling of shell arguments.
Audit Metadata