The vulnerability in NousResearch hermes-agent arises from improper handling of input in the detect_dangerous_command function within the terminal_tool component, specifically in tools/approval.py. This flaw enables remote attackers to perform OS command injection, potentially allowing execution of arbitrary commands on the affected system. The affected version is identified by the commit hash 5157f5427f19488b31c6fdebbacd15d798ce7f63. Despite early vendor notification, no official fix or advisory has been issued. The vulnerability is publicly disclosed, increasing the risk of exploitation, although no active exploitation has been documented.

Successful exploitation could allow remote attackers to execute arbitrary operating system commands on the affected device, potentially leading to unauthorized system control or data compromise. The CVSS score of 6.9 reflects a medium severity with network attack vector and no required privileges or user interaction. However, the lack of an official patch or mitigation increases the risk for affected users.

Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. Since the vendor has not responded and no official fix is available, users should consider applying manual mitigations such as restricting network access to the hermes-agent service, employing application-level input validation or sandboxing, and monitoring for suspicious activity related to command execution. Stay alert for any future vendor advisories or community patches.