hermes-agent
Pass
Audited by Gen Agent Trust Hub on May 10, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The documentation describes an installation process using a one-line shell script fetched from the official Nous Research GitHub repository (
curl ... | bash). This is the standard deployment method for the tool. - [COMMAND_EXECUTION]: The skill provides a detailed reference for CLI commands that interact with the host system, including file management, background process handling, and service installation. It explicitly documents security boundaries such as Docker isolation and dangerous command approval flows to mitigate risks.
- [PROMPT_INJECTION]: Several documentation files list common prompt injection strings (e.g., "ignore previous instructions") as examples of patterns that the agent's internal security scanner is designed to detect and block. These are documented for educational purposes and do not represent active injection attempts.
- [EXTERNAL_DOWNLOADS]: The guides mention fetching Whisper models for speech-to-text and installing various Python and Node.js dependencies required for specific features like voice mode or browser automation.
Audit Metadata