Unauthenticated Remote Code Execution in AutoAgent TCP Server

AutoAgent features a critical unauthenticated remote code execution vulnerability (CVE-2026-86124) within its TCP server component. The TCP server is configured to bind to all available network interfaces, leaving it exposed to external connection attempts. An attacker can connect directly to the service port and send input that the server executes as root within the container environment. This flaw provides a direct path for command injection, enabling full control over the container and access to any bind-mounted host workspace directories. Given the service's default configuration of binding to all interfaces and executing commands with root privileges, this vulnerability poses an immediate and high-impact risk to any environment running an exposed instance of AutoAgent.

Impact

Successful exploitation results in full remote code execution with root privileges within the AutoAgent container. This allows attackers to compromise the container integrity, exfiltrate sensitive data from bind-mounted host workspace directories, and potentially use the container as a pivot point for further lateral movement within the host infrastructure.

Recommendation

• Immediately restrict network access to the AutoAgent TCP server port using host or network firewalls to ensure it is not reachable from untrusted interfaces.
• Monitor network traffic for unauthorized connections to ports typically utilized by the AutoAgent TCP service.
• Audit container configurations to identify and remove unnecessary bind-mounted host volumes to limit the scope of potential post-exploitation access.