The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using yt-dlp and ffmpeg. While the examples wrap the URL placeholder in double quotes (e.g., yt-dlp "URL"), there is a risk of command injection if the agent does not properly sanitize shell metacharacters in the user-supplied URL.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external websites via video metadata.
Ingestion points: Data enters the agent's context through the --dump-json command, which retrieves video titles, descriptions, and other metadata from external URLs (SKILL.md).
Boundary markers: None provided; there are no instructions to the agent to treat the JSON output as untrusted or to ignore embedded instructions.
Capability inventory: The skill uses shell command execution via yt-dlp and ffmpeg and has the capability to write files to the local file system (media downloads).
Sanitization: No sanitization or validation of the retrieved metadata is mentioned before processing.

video-download