video
Pass
Audited by Gen Agent Trust Hub on Jul 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill includes instructions to run standard installation commands for video production frameworks, such as
npm install hyperframesandnpx create-video@latest. These are documented uses for well-known and legitimate programmatic video libraries. - [PROMPT_INJECTION]: The skill describes a 'Reverse-Engineering' workflow where the agent analyzes external video content or social media posts via tools like
watch-video. This constitutes an ingestion point for indirect prompt injection. However, the risk is mitigated by an explicit 'Review once, then execute' instruction, requiring a manual beat-sheet approval before any content is generated, and a strict guardrail to only copy editing grammar rather than content.
Audit Metadata