general-video
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill directs the agent to run various local utility scripts using Node.js, such as audio-discover.cjs, check-fast-passages.cjs, and audio.mjs, and to use the HyperFrames CLI tools (npx hyperframes) for linting and validation of the video project.
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided production briefs and design specifications (frame.md, design.md), which represents an ingestion surface for data that influences the agent's video composition and animation behavior.
- Ingestion points: User production prompts, design identity files, and audio request metadata.
- Boundary markers: Employs a mechanism to force manual confirmation of visual identity and planning decisions before technical implementation.
- Capability inventory: Execution of shell commands (node, npx), local filesystem read/write access, and media rendering via FFmpeg.
- Sanitization: The process relies on manual oversight and a review-driven production pipeline rather than automated sanitization of input briefs.
Audit Metadata