Steering Prototypes with Prompt-tuning for Rehearsal-free Continual Learning