video-ads
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes several scripts (
package.sh,check.mjs,rw.mjs) to automate video production and verification. These scripts execute standard multimedia tools includingffmpeg,ffprobe, and theremotionCLI (via npx). - [EXTERNAL_DOWNLOADS]: The skill facilitates communication with the Runway AI API to generate video footage. Additionally, it downloads standard project dependencies from the NPM registry, fonts from Google Fonts, and a headless Chrome environment required for rendering via the Remotion framework.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external inputs such as project briefs, news stories, and transcripts. To mitigate potential injection or safety risks, the skill implements a strict 'Hard Gate' requiring a human decision-owner to manually type an 'APPROVED' marker in the project spine before any assets are generated or built. It also enforces the use of a
SOURCES.mdfile to verify all on-screen claims. - [SAFE]: The skill incorporates exemplary safety practices for AI agent skills, including a credit ledger system (
rw.mjs) to prevent unauthorized or accidental overspending on AI generation services, and a comprehensive 'Publication Gate' checklist that covers rights, consent, and platform classification.
Audit Metadata