I wonder if adding another agent that make code review (Create loop of implementation and code review fixes. one agent implement task, second agent give him rejects) before bring it to humen code review will be beneficial.
Very likely. I've been planning to add that along with general code code improvement workflows.