AutoJack MCP Loopback Trust Abuse: autogenstudio Spawning Shell Processes

Ibrahim Saud@tektrix

•updated Jun 28, 2026•0•0•2

This rule detects instances where AutoGen Studio or related AI agents initiate potentially suspicious child processes such as command interpreters (cmd, powershell, bash) or known Living-off-the-Land Binaries (LOLBins) while simultaneously establishing outbound WebSocket connections to Model Context Protocol (MCP) endpoints. This behavior is indicative of an attacker attempting to leverage AI agent automation capabilities for code execution or remote system interaction.