The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches versioned JavaScript libraries (such as MapLibre and TopoJSON) and geographic boundary data from public CDNs like jsdelivr.net. It also utilizes npx hyperframes add to download animation blocks from its ecosystem registry.
[COMMAND_EXECUTION]: The workflow involves multiple shell commands executed via npx for project initialization, asset sourcing, and media rendering. Additionally, it spawns ffmpeg and ffprobe as subprocesses for video encoding and image metadata retrieval.
[DYNAMIC_EXECUTION]: The map rendering process uses Puppeteer to launch a headless browser and execute dynamic JavaScript within a generated HTML page. This process uses the --no-sandbox flag, which is a documented choice for compatibility in CI/CD environments but reduces isolation when processing web content.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and animate data from external sources, such as news articles, tweets, and web pages, creating a potential surface for malicious instructions embedded in that data.
Ingestion points: Untrusted data enters the workflow via hyperframes capture for web pages and via automated searches for news and social media content.
Boundary markers: The system uses predefined layout templates and CSS-based animation rules to process data, which acts as a structural boundary.
Capability inventory: The toolset includes the ability to execute shell commands, perform network fetches through a browser, and write media files to the local disk.
Sanitization: The documentation indicates an awareness of sanitization, specifically mentioning it as a requirement when importing assets from external design tools.

motion-graphics