The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run to execute the yt-dlp command-line tool. The implementation in scripts/download.py correctly constructs the command as a list of arguments, which is a secure practice that prevents shell injection vulnerabilities even when handling untrusted URLs or user-defined output paths.
[EXTERNAL_DOWNLOADS]: The skill is designed to download media content from external video platforms. It also provides instructions for installing the yt-dlp tool and the bgutil-ytdlp-pot-provider plugin from standard, well-known registries (PyPI, Homebrew).
[INDIRECT_PROMPT_INJECTION]: The skill has a potential attack surface for indirect prompt injection as it processes external URLs.
Ingestion points: Untrusted video URLs passed as arguments to the download.py script and SKILL.md instructions.
Boundary markers: Not explicitly defined in the agent instructions.
Capability inventory: The skill can write files to the local file system (defaulting to ~/Downloads/Videos) and perform network requests via the yt-dlp subprocess.
Sanitization: Input URLs are passed as literal strings in a subprocess list, and output paths are resolved using Path.resolve(), mitigating command injection risks.

download-video