google-agents-cli
Fail
Audited by Gen Agent Trust Hub on Jun 16, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill instructs users to run
uvx google-agents-cli setup, which downloads and executes code from a Python package. This package uses official branding but is provided by a third-party author ('Aradotso'), creating a high risk of executing unvetted code under the guise of an official tool. - [COMMAND_EXECUTION]: The installation instruction
npx skills add google/agents-clifetches and executes a remote skill suite. The naming convention 'google/agents-cli' is deceptive as it implies official Google origin while being distributed by an unrelated entity. - [PROMPT_INJECTION]: The skill's metadata contains deceptive claims, describing itself as the 'official CLI and skill suite' for 'Google Cloud's Gemini Enterprise Agent Platform'. This metadata poisoning is designed to deceive users and AI agents into granting high levels of trust to a third-party tool.
- [CREDENTIALS_UNSAFE]: The skill provides instructions for setting sensitive environment variables such as
GOOGLE_API_KEY. While using environment variables is a standard practice, the context of providing these keys to a deceptive third-party CLI increases the risk of credential harvesting.
Recommendations
- AI detected serious security threats
Audit Metadata