google-veo
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of user-provided text prompts to generate video content using external models. This represents an entry point for untrusted data into the agent's workflow.\n
- Ingestion points: User-supplied prompts are interpolated into the JSON input for the
belt app runcommand withinSKILL.mdexamples.\n - Boundary markers: Prompts are encapsulated within structured JSON objects passed to the CLI.\n
- Capability inventory: The skill utilizes the
beltCLI via theBashtool to perform network requests to the inference service.\n - Sanitization: The skill relies on the underlying
belttool and theinference.shplatform to handle and sanitize input before processing.\n- [EXTERNAL_DOWNLOADS]: The skill references external resources for installation and configuration, pointing to the service provider's infrastructure.\n - Fetches installation instructions from
raw.githubusercontent.com/inference-sh/skills/.\n - Directs users to official documentation on
inference.shdomains.
Audit Metadata