autogen

This module itself contains no obvious malicious code, secrets, or direct data-stealing logic. However, it implements a fully automated pipeline where untrusted LLM output is used to drive code execution inside a Docker executor (human_input_mode="NEVER"), creating a meaningful security risk. The true risk level depends on the Docker executor’s sandboxing (network/file/privilege restrictions) and any policy enforcement in the wider codebase, which are not shown here.