google-veo
Pass
Audited by Gen Agent Trust Hub on Mar 27, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads an installation script from
https://cli.inference.sh. This is the official domain for the skill vendor's command-line interface. - [REMOTE_CODE_EXECUTION]: The instructions provide the command
curl -fsSL https://cli.inference.sh | shto install theinfshutility. This pattern downloads and executes a script directly from the vendor's infrastructure. - [COMMAND_EXECUTION]: The skill utilizes the
infshCLI via theBashtool to perform operations such as logging in and running video generation models (e.g.,infsh app run). - [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface by interpolating user-provided text into a command-line tool's arguments.
- Ingestion points: User-supplied prompt strings are processed via the
--inputJSON parameter inSKILL.md. - Boundary markers: The input is contained within a JSON object and enclosed in single quotes for shell execution.
- Capability inventory: The skill executes shell commands via the
Bashtool to interact with the vendor's API. - Sanitization: There is no explicit sanitization described for the prompt input, relying on the underlying tool for input handling.
Audit Metadata