The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external, untrusted video and audio content and convert it into text transcripts. This introduces a surface for indirect prompt injection, where malicious instructions hidden within the audio or metadata of a video could be parsed and followed by the agent during subsequent processing steps.
Ingestion points: Audio and video files sourced from arbitrary external URLs provided by the user.
Boundary markers: Absent. The transcript generated by scripts/transcribe_siliconflow.py is written to a file and presented to the agent without specific delimiters or instructions to ignore embedded commands.
Capability inventory: The skill can execute shell commands, perform network requests to an external API (SiliconFlow), and write files to the local file system.
Sanitization: There is no evidence of sanitization or filtering of the transcribed text before it is presented to the agent or saved to disk.
[COMMAND_EXECUTION]: The skill documentation encourages the use of the yt-dlp flag --cookies-from-browser. This functionality allows the tool to read and extract authentication cookies from local browser profiles (e.g., Chrome, Edge). While necessary for downloading age-restricted or member-only content, this capability grants the agent indirect access to the user's sensitive session data and credentials.

advanced-video-downloader