The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted video links and sharing codes from multiple social media platforms.
Ingestion points: The Input section of SKILL.md specifies video links from platforms like Douyin, YouTube, and Bilibili as primary inputs.
Boundary markers: There are no instructions to use delimiters or ignore potential control sequences within the processed content.
Capability inventory: The agent is instructed to use yt-dlp, ffmpeg, and Chrome via shell commands to process these inputs.
Sanitization: The skill does not provide instructions for validating or escaping URLs before they are passed to system commands.
[COMMAND_EXECUTION]: The skill requires the execution of several system binaries with external arguments.
Evidence: Instructions specify the use of yt-dlp for downloading, ffmpeg for extraction, and ffprobe for metadata collection.
Evidence: The fallback mechanism involves launching chrome.exe with the --remote-debugging-port=9223 and --no-proxy-server flags to automate browser actions via CDP.
[CREDENTIALS_UNSAFE]: The skill documentation suggests the use of browser cookies to bypass platform restrictions.
Evidence: The workflow mentions that yt-dlp typically requires browser cookies to retrieve high-definition sources from certain platforms, which necessitates the agent accessing sensitive session data.
[EXTERNAL_DOWNLOADS]: The skill's primary function is fetching content from various external remote domains.
Evidence: It provides instructions to download media from a wide range of social media and video hosting platforms using automated tools.

rap-clone-s1-download