video
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to analyze external video and social media content to reverse-engineer editing grammar and rhythm. This creates an untrusted data ingestion surface.
- Ingestion points: Uses the
watch-videoandsocial-fetchtools to pull external media content as defined inreferences/edit-anatomy.md. - Boundary markers: The skill requires the creation of a 'beat sheet' which forces the agent to map visual patterns into a structured technical table, helping to isolate the analysis of form from the execution of embedded content.
- Capability inventory: The skill allows for local code generation (HTML/CSS/JS via Hyperframes and React via Remotion) and remote video generation via APIs.
- Sanitization: Relies on the inherent safety guardrails of the multimodal tools and the agent's internal content filtering mechanisms.
- [EXTERNAL_DOWNLOADS]: The skill references standard package installation and execution flows (
npm install hyperframes,npx create-video@latest). These utilities belong to well-known open-source frameworks and trusted service providers in the video technology space.
Audit Metadata