google-veo
Fail
Audited by Gen Agent Trust Hub on Apr 4, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches the CLI installation script from the vendor's infrastructure at
https://cli.inference.sh.- [REMOTE_CODE_EXECUTION]: Executes the downloaded installation script by piping it directly to the shell (curl ... | sh).- [COMMAND_EXECUTION]: Utilizes theinfshtool to perform video generation and platform management tasks.- [PROMPT_INJECTION]: Contains a vulnerability surface for indirect prompt injection where untrusted data is processed. \n - Ingestion points: User-provided video prompts are interpolated into the
--inputargument of shell commands inSKILL.md.\n - Boundary markers: Prompts are enclosed within a JSON object string inside single quotes in the command line.\n
- Capability inventory: The skill has permission to execute shell commands via
Bash(infsh *)as defined in the YAML frontmatter.\n - Sanitization: There is no evidence of input validation or escaping for the prompt content before it is used in the command line.
Recommendations
- HIGH: Downloads and executes remote code from: https://cli.inference.sh - DO NOT USE without thorough review
Audit Metadata