Meta-prompting

Meta-prompting uses a Large Language Model (LLM) to dynamically generate, refine, or optimize the instructions (prompts) for a subsequent LLM task.

This advanced technique shifts prompt engineering from a manual task to an autonomous process: you instruct the AI to become its own prompt engineer. The LLM first creates a structured, high-quality prompt (the 'meta-prompt') based on your initial, simple request (e.g., 'Write code to process JSON data'). This self-refinement loop ensures the final prompt includes critical components (context, constraints, error handling), leading to demonstrably more accurate and robust outputs. In complex workflows, a 'conductor' LLM can use meta-prompting to orchestrate multiple specialist LLMs, improving overall system efficiency and performance consistency across varied, multi-step tasks.

https://www.promptingguide.ai/techniques/meta-prompting