The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on executing shell commands for yt-dlp, ffmpeg, and ffprobe. These commands use arguments such as URLs and file paths provided by the user or fetched from the environment.
[DATA_EXFILTRATION]: The skill utilizes the --cookies-from-browser flag, which grants access to sensitive authentication data stored in the user's browser. Although this is used for authorized content access, it represents a data exposure surface. The instructions include specific prohibitions against displaying, exporting, or committing cookie data to the repository.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from YouTube video metadata (titles, descriptions, and captions) when it writes and reads .info.json files.
Ingestion points: Reads metadata and caption files downloaded from YouTube via yt-dlp in SKILL.md.
Boundary markers: The instructions do not specify formal delimiters for the processed metadata, but they do include a directive for the agent to treat titles and descriptions as unverified evidence rather than proof.
Capability inventory: The skill has the ability to write files to the local system and execute subprocesses (yt-dlp, ffprobe).
Sanitization: The skill recommends labeling automatic captions to distinguish them from human-verified transcripts and cross-referencing metadata with other speaker-naming skills.

edgespeak-yt-download