A prompt tells an AI what task to attempt, what context matters and what kind of answer will be useful.
Do now: Rewrite one vague request with a task, audience and output format.
A prompt guides the attempt. It does not guarantee the result.Prompts are the interface between your intention and a model’s output. Clearer instructions reduce guessing, make review easier and help you reuse good work.
Prompt means input, not magic
A prompt can be a question, command, example, image, document or combination of inputs. The model uses it with learned patterns and available tools to produce an output.
The same prompt can produce different results across tools, model versions and settings. Judge the output rather than assuming the wording has permanent power.
Good prompting makes your intent testable.
Name the work and the reader
Begin with a verb: compare, summarise, draft, classify, explain or question. Then name the person who will use the answer and the decision it should support.
Add only context that changes the answer. If a fact is missing, invite a question rather than letting the system invent one.
Describe a usable return
Request a format that fits the next step: table, checklist, email, outline or JSON. Include length, tone or headings only when they serve the user.
Finish with a check: cite sources, label assumptions, preserve exact figures or flag uncertainty. Then review the output yourself.
The best prompt is the smallest instruction that reliably produces a usable draft.