Meta-prompting.
THE TECHNIQUE
Ask the model to improve your prompt before running it.
"Here is a prompt I plan to use. What is unclear or missing?"
WHY IT HELPS
It identifies missing context, ambiguity and unstated assumptions.
Frequently it asks the question you should have answered.
USEFUL VARIANTS
"What would you need to know to do this well?" "Rewrite this prompt to be clearer." "What is ambiguous here?"
THE FIRST ONE
Particularly useful. The model lists what it is missing, and you supply it.
FOR BUILDING REUSABLE PROMPTS
Draft it, test it, ask what failed, and fold the answer back in.
After several rounds it produces output needing little editing.
WHAT TO BE CAREFUL WITH
Suggested prompts that are longer without being better.
Length is not quality.
WHEN THIS IS WORTH DOING
For prompts you will reuse.
Not for a one-off question, where the effort exceeds the benefit.
