hermes-web-ui-dashboard
Fail
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONPERSISTENCEPRIVILEGE_ESCALATION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill documentation includes an "Auto-Setup Script" for Linux and macOS that downloads a shell script from a remote GitHub repository and executes it directly using bash process substitution:
bash <(curl -fsSL https://raw.githubusercontent.com/EKKOLearnAI/hermes-web-ui/main/scripts/setup.sh). This allows for arbitrary code execution on the host machine from an untrusted source. - [COMMAND_EXECUTION]: The dashboard includes an integrated "Web Terminal" feature that provides a full PTY (pseudoterminal) interface over WebSockets. This allows for arbitrary command execution on the server hosting the dashboard.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and manage chat sessions from external platforms like Telegram, Discord, Slack, and WhatsApp. This data is untrusted and is processed by the AI agent, creating a vulnerability surface where malicious inputs from these platforms could potentially override agent instructions or trigger unauthorized tool usage.
- [PERSISTENCE]: The documentation provides explicit instructions for creating a
systemdservice (/etc/systemd/system/hermes-web-ui.service) to ensure the application starts automatically on boot and maintains persistence. - [PRIVILEGE_ESCALATION]: The instructions suggest using
sudofor file ownership management (chown) and for managing system services (systemctl), which could be leveraged if the environment is not properly isolated.
Recommendations
- HIGH: Downloads and executes remote code from: https://raw.githubusercontent.com/EKKOLearnAI/hermes-web-ui/main/scripts/setup.sh - DO NOT USE without thorough review
- AI detected serious security threats
Audit Metadata