The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes system tools to process and render media. It executes ffmpeg to encode video frames into MP4 and GIF formats, and uses curl to download image assets. It also uses npm to install required dependencies and provides a utility to generate an asset-locking bash script.
[EXTERNAL_DOWNLOADS]: The skill fetches resources from well-known services and organizations. It downloads public domain image metadata and assets from NASA, The Metropolitan Museum of Art, and the Art Institute of Chicago. It also downloads a Chromium browser build via the Playwright library to handle canvas rendering.
[INDIRECT_PROMPT_INJECTION]: The skill represents a potential injection surface as it ingests metadata from external museum APIs and processes user-supplied configuration files.
Ingestion points: scripts/source.mjs (external API responses), scripts/render.mjs (local project configuration and image files).
Boundary markers: None present.
Capability inventory: Video encoding via ffmpeg, local HTTP server for asset rendering, and file system writes for project asset locking.
Sanitization: The skill implements directory traversal protection in its internal asset server to prevent unauthorized file access and strips HTML from external metadata.

horizon-reel