install-skills
Fail
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: HIGHCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes several commands on a remote host via SSH, including 'hermes-setup check', 'hermes-setup catalog', and 'hermes-setup add'.
- [REMOTE_CODE_EXECUTION]: The skill installs and potentially executes code from the 'ai-maxxing' package on the user's VPS. The installation process involves fetching external components and placing them in the '~/.agents/skills' directory, which is then used by the agent.
- [INDIRECT_PROMPT_INJECTION]: The skill accepts user-supplied arguments (skill names) and interpolates them directly into a shell command: 'ssh hermes 'hermes-setup add <base|all|skill>...''. Since these inputs are not sanitized, a malicious user or an indirect injection could provide a payload (e.g., '; rm -rf /') to execute arbitrary commands on the remote server.
Recommendations
- AI detected serious security threats
Audit Metadata