A framework to move beyond simple prompting towards defining *how* the LLM should structure its internal processing, access its latent knowledge, and apply specific heuristics or constraints when dealing with a particular subject matter or task.