The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes content from external video platforms, which can contain untrusted data in metadata fields such as titles and descriptions.
Ingestion points: Video metadata (titles, descriptions, thumbnails) and playlist content fetched from external URLs provided by the user as described in SKILL.md.
Boundary markers: Absent. The skill does not provide instructions to the agent to ignore or delimit potentially malicious instructions found within video metadata.
Capability inventory: File system writing to ~/Downloads/ and shell command execution via the yt-dlp and FFmpeg utilities.
Sanitization: Absent. There is no mention of filtering or sanitizing metadata before it is processed or saved to the filesystem.
[COMMAND_EXECUTION]: The skill requires the use of CLI tools (yt-dlp, FFmpeg) to perform its primary function. While these are standard tools, the agent must execute shell commands to use them. If the agent does not properly sanitize external inputs like video URLs or filenames derived from video titles, it could be vulnerable to command injection.

video-downloader