youtube-video-to-skill
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill relies on several shell commands to function, including
pip install,yt-dlp,ffmpeg, anduvx. These are used to manage dependencies, download media, and run transcription models. - [DATA_EXPOSURE]: The skill instructs the use of
yt-dlp --cookies-from-browser chrome. This command attempts to read the user's Chrome browser database to extract session cookies. These cookies are sensitive as they can be used to hijack active web sessions and bypass authentication. - [EXTERNAL_DOWNLOADS]: The skill downloads the
yt-dlppackage and a machine learning model (whisper-large-v3-turbo) from the mlx-community repository on Hugging Face at runtime. - [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core functionality.
- Ingestion points: The agent transcribes external YouTube videos (SKILL.md). A transcript could contain hidden instructions intended to manipulate the agent's behavior during the mapping and generation steps.
- Boundary markers: No delimiters or instructions are used to separate the transcript data from the agent's logic.
- Capability inventory: The agent has capabilities to execute shell commands, run Python code, and write files to the local repository (SKILL.md).
- Sanitization: No validation or sanitization is performed on the transcribed text before it is processed by the agent.
Audit Metadata