The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run to invoke the yt-dlp CLI and various helper scripts (scripts/doctor.py, scripts/probe_url.py, scripts/check.py). These calls use list-based arguments, preventing shell injection vulnerabilities from user-supplied URLs or paths.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, specifically video metadata and transcripts retrieved from remote URLs via yt-dlp.
Ingestion points: Metadata is retrieved as JSON by scripts/probe_url.py, and subtitles are fetched as text files by the main CLI.
Boundary markers: The instructions do not explicitly mandate delimiters when the agent summarizes transcripts or metadata for the user.
Capability inventory: The skill has access to the Bash and Write tools, allowing it to execute commands and write to the filesystem.
Sanitization: The Python scripts parse the remote data using standard JSON libraries, minimizing the risk of parsing-level exploits.

yt-dlp