motion-ad
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes various system commands to perform its duties, including
npm installfor dependency management,curlfor downloading web assets,ffmpegfor video rendering, andpython3for image processing. - [EXTERNAL_DOWNLOADS]: Assets are fetched from user-provided URLs. The script
scripts/fetch-images.cjsautomates the discovery and download of images from a landing page to ensure the ad uses authentic product imagery. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon data from external websites.
- Ingestion points: Website content is processed in
scripts/fetch-images.cjsto extract images and in the workflow to verify claims. - Boundary markers: The instructions specify that on-screen claims must match the source page and that only real assets should be used, providing a framework for the AI to follow.
- Capability inventory: The skill can execute
curl,ffmpeg, and perform file system operations. - Sanitization: URLs are validated through standard library parsers, and downloaded files are assigned sanitized, numerically indexed names.
Audit Metadata