The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the Bash tool to execute yt-dlp commands with user-supplied arguments.
Evidence: Multiple instruction blocks in SKILL.md demonstrate executing commands like yt-dlp -F "URL" and yt-dlp -o "%(title)s.%(ext)s" "URL".
Risk: If the user provides a URL containing shell metacharacters (e.g., ;, &&, $()), it could lead to arbitrary command execution if the agent does not properly escape the input.
[EXTERNAL_DOWNLOADS]: The skill's primary purpose is to download media content from external servers.
Evidence: Explicitly supports downloads from YouTube, Vimeo, Twitter, TikTok, and other platforms using yt-dlp.
Context: This behavior is the stated primary purpose of the skill.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection via the processing of untrusted URLs.
Ingestion points: The URL and PLAYLIST_URL parameters in SKILL.md act as entry points for untrusted data into the shell environment.
Boundary markers: While the examples use double quotes around the URL placeholder, this does not prevent all forms of command substitution in a shell context.
Capability inventory: The skill is granted Bash tool access, which can be misused if input is injected into commands.
Sanitization: There are no explicit instructions for the agent to validate or sanitize the URL format or content before passing it to the Bash tool.

claude-video-download