The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to drive the yt-dlp and ffmpeg CLI tools. It provides defensive patterns such as mandatory quoting of every URL in shell commands to prevent command injection vulnerabilities.
[EXTERNAL_DOWNLOADS]: The skill facilitates downloading media from external streaming sites and includes instructions for using the official yt-dlp -U self-update mechanism to maintain the tool's effectiveness against site changes.
[INDIRECT_PROMPT_INJECTION]: The skill handles untrusted data in the form of streaming-site URLs.
Ingestion points: External media URLs are ingested into the terminal tool via yt-dlp commands in SKILL.md.
Boundary markers: Explicit instructions require quoting all URLs to maintain shell command boundaries.
Capability inventory: The agent can execute commands, write files to disk, and process media using ffmpeg as defined in SKILL.md.
Sanitization: The skill relies on shell quoting and the internal validation mechanisms of the yt-dlp binary to handle external inputs.
[CREDENTIALS_UNSAFE]: While the skill involves handling authentication mechanisms like cookies and PO tokens, it implements strong security guidance by forbidding the agent from logging or exposing these secrets and requiring explicit user approval for local browser data access.

yt-dlp-cli