youtube-transcription
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the user to install well-known and trusted Python packages: 'openai-whisper' and 'yt-dlp'. These are legitimate tools for transcription and video processing.
- [COMMAND_EXECUTION]: The skill uses shell commands to download audio from YouTube URLs using 'yt-dlp' and run transcription using the 'whisper' CLI. While these commands process external input (URLs and audio files), they are executed within the scope of the skill's primary purpose.
- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8c) because it reads and processes data from an external, attacker-controlled source (YouTube). A malicious video could contain audio or metadata designed to influence the AI's behavior when it processes the resulting transcript. The risk is assessed as LOW due to standard LLM guardrails and the nature of the tool.
Audit Metadata