google-veo
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute commands for the
beltCLI, enabling video generation and model management. The usage is constrained to thebeltapplication through the tool's configuration. - [EXTERNAL_DOWNLOADS]: The skill documentation references external resources, including installation guidelines and supplemental skills from the
inference-shGitHub organization, which are required for the intended functionality. - [INDIRECT_PROMPT_INJECTION]: 1. Ingestion points: User-defined text prompts are passed to the
belt app runcommand inSKILL.md. 2. Boundary markers: The input is encapsulated within a JSON object structure. 3. Capability inventory: The skill executes shell commands via theBashtool. 4. Sanitization: No explicit filtering or sanitization of the input strings is mentioned in the instructions.
Audit Metadata