The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the yt-dlp CLI tool to download media, extract audio, and fetch metadata from user-provided URLs. Commands are structured to use double quotes around the URL variable to mitigate basic shell injection risks.
[EXTERNAL_DOWNLOADS]: Recommends installing yt-dlp and ffmpeg using official and well-known package managers such as pip, brew, and apt.
[REMOTE_CODE_EXECUTION]: Documents the use of yt-dlp -U and pip install -U yt-dlp, which are the official self-update mechanisms for the tool to maintain compatibility with media hosting sites.
[INDIRECT_PROMPT_INJECTION]: The skill provides a surface for processing untrusted user input via the URL parameter.
Ingestion points: The "URL" placeholder in the provided yt-dlp commands (e.g., SKILL.md).
Boundary markers: The skill instructions correctly wrap the input variable in double quotes to prevent immediate shell metacharacter execution.
Capability inventory: Includes network requests, file system writes (media downloads), and metadata extraction via subprocess execution.
Sanitization: Relies on the agent's shell quoting and yt-dlp's internal handling of remote URLs.

video-download