The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run in scripts/download_video.py to execute yt-dlp. Commands are passed as lists of arguments, which prevents shell-based command injection attacks.
[EXTERNAL_DOWNLOADS]: The script scripts/download_video.py automatically installs the yt-dlp package from PyPI using pip if it is not found on the system. yt-dlp is a standard, widely-used tool for this purpose.
[PROMPT_INJECTION]: The skill processes untrusted data from YouTube (video metadata) which could be used for indirect prompt injection.
Ingestion points: Metadata is ingested through yt-dlp --dump-json and parsed in scripts/download_video.py.
Boundary markers: None are present to separate untrusted metadata from instructions when displaying video information.
Capability inventory: The skill has file system access (writing to /mnt/user-data/outputs/) and subprocess execution capabilities.
Sanitization: There is no sanitization of video titles or uploader information before it is processed or printed.

video-downloader