dor4.net

← Guides

Write better prompts

A repeatable way to give AI systems clear goals, useful context, and checkable requirements.

Good prompting is less about finding a magic phrase and more about writing a clear brief. A strong prompt makes the desired outcome, relevant context, and constraints easy to distinguish.

Start with the outcome

State what should exist when the work is finished. “Help with this report” leaves the task open. “Turn these notes into a one-page decision memo for a technical manager” defines an observable result.

Add the reason when it changes the answer. A summary for a decision, a lesson, and a public announcement require different emphasis even when they use the same source material.

Supply only relevant context

Include facts the model cannot safely infer: the audience, existing decisions, terminology, examples, and source material. Separate background information from instructions so that quoted text is not mistaken for a command.

Large context is not automatically useful. Remove duplicated notes and mark the sections that deserve the most attention.

Make constraints explicit

Useful constraints include:

Prefer concrete rules over vague adjectives. “Use four short sections and avoid marketing language” is easier to follow than “make it professional.”

Ask for a checkable structure

Request output you can inspect: a table with named columns, a numbered plan, a patch plus tests, or a recommendation with assumptions and trade-offs. Structure reduces ambiguity and makes omissions visible.

Iterate on evidence

If the first result misses the mark, identify the specific failure. Add the missing constraint or correct the misunderstood fact. Avoid restarting with a completely different prompt unless the goal itself changed.

A compact template is often enough:

Produce [outcome] for [audience]. Use [context]. Follow [constraints]. Return [format]. A successful result must [acceptance criteria].