Use when the user invokes /goal, asks for goal mode, designs an agent loop, or starts long-horizon autonomous work that should be clarified into a concrete, verifiable objective before execution.
