url-to-motion-design
Pass
Audited by Gen Agent Trust Hub on Nov 11, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches video content and metadata from external sources including X (Twitter) via api.fxtwitter.com, YouTube, and Vimeo using the yt-dlp utility. These operations are core to the skill's primary function of processing video URLs.
- [COMMAND_EXECUTION]: Multiple scripts (scripts/fetch-source.mjs, scripts/shots.mjs, scripts/render.mjs, scripts/video.mjs) execute external CLI tools such as ffmpeg, ffprobe, magick (ImageMagick), and yt-dlp. These are used for video extraction, frame rendering, and media encoding.
- [DYNAMIC_EXECUTION]: The server and screening scripts dynamically import JavaScript modules and read assets from a local repository path (facet-studio) configured via environment variables or expected directory structures. Additionally, the lab server dynamically assembles HTML pages for rendering in a sandboxed iframe using a template system.
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting untrusted metadata (titles, descriptions, and post text) from external video platforms.
- Ingestion points: scripts/fetch-source.mjs extracts text content from video metadata and X posts into source.json, which is later read by helper agents in SKILL.md.
- Boundary markers: No explicit delimitation or 'ignore embedded instructions' warnings are applied to the ingested metadata before it is processed by builder agents.
- Capability inventory: The skill environment has capabilities for file system writes, subprocess execution (ffmpeg), and network access.
- Sanitization: There is no evidence of sanitization or filtering of the extracted video metadata.
Audit Metadata