veo-video
Pass
Audited by Gen Agent Trust Hub on Apr 8, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Downloads generated video media and reference images from Google's official Gemini API infrastructure (generativelanguage.googleapis.com).
- [DATA_EXFILTRATION]: Transmits user-provided text prompts and local image data to Google's generative AI services for processing. The authentication token (GEMINI_API_KEY) is sent to established Google API endpoints.
- [PROMPT_INJECTION]: The skill acts as a surface for indirect prompt injection as it processes untrusted user input that is then passed to a large language model.
- Ingestion points: User-supplied strings via the
--promptand--image-promptCLI arguments, and local files via--image. - Boundary markers: None; inputs are interpolated directly into the JSON request body sent to the API.
- Capability inventory: Performs file system read/write operations (
fs.readFileSync,fs.writeFileSync) and network requests (fetch). - Sanitization: Input prompts are not sanitized or filtered before being transmitted to the external model.
Audit Metadata