The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches reference thumbnail images from YouTube's image servers and video metadata to inform the design process during the research phase.- [EXTERNAL_DOWNLOADS]: It instructs the agent to search for and download official brand logos from external web sources identified through user-guided search results to use as graphic assets.- [COMMAND_EXECUTION]: Uses the yt-dlp utility to scrape search results, view counts, and video titles directly from YouTube.- [COMMAND_EXECUTION]: Executes a series of image manipulation commands via ImageMagick (magick, convert, composite) to post-process generated images and apply branding.- [COMMAND_EXECUTION]: A Python script is executed at runtime to decode Base64 image data received from API responses and write it to the local file system.- [DATA_EXFILTRATION]: Transmits local subject image data and design prompts to Google's generative language API for thumbnail creation, utilizing the GEMINI_API_KEY environment variable for authentication.- [PROMPT_INJECTION]: The skill processes untrusted external data such as YouTube video titles and logo metadata, which constitutes an indirect prompt injection surface.
Ingestion points: Video titles and search metadata retrieved via yt-dlp and external logo search results.
Boundary markers: No specific delimiters or warnings are used to separate external content from internal instructions.
Capability inventory: The skill possesses capabilities for shell execution, network requests, and file system manipulation.
Sanitization: No explicit sanitization or filtering of the retrieved external metadata is performed before the data influences subsequent agent actions.

thumbnail