While large language models have been touted for their potential in cybersecurity, they are still far from executing real-world cyber attacks — unless given help from a new kind of abstraction layer, ...