A benchmark for evaluating how well AI coding agents can cooperate on software engineering tasks with potential conflicts.