A prompt-cache countdown in the prompt footer, beside the model and effort: how long until your next message has to re-cache the whole conversation.