Prompt builder

How to use it

1. Optionally pick a template to fill the fields with a good starting point.
2. Edit each field for your task. Leave any field empty to skip it.
3. Choose a layout style and copy the finished prompt.

What makes a prompt work

Models are capable but not mind-readers. Most disappointing answers come from prompts that leave out something a human expert would ask about first: who is this for, what is it for, what should it look like, and what should be avoided. The builder breaks a prompt into those parts so nothing important is missed.

The fields

• Role. The expertise and viewpoint you want, such as "an experienced property investment analyst in Malaysia". This sets vocabulary and depth.
• Task. What you want done, stated as a clear instruction.
• Context. Background the model cannot know: your business, the situation, the source material.
• Audience. Who will read the result and what they already know.
• Constraints. Rules to follow: length, tone, things to include or avoid, facts not to invent.
• Output format. The shape of the answer: bullet list, table, email, JSON with specific fields.
• Example. A sample of a good output. Models copy the structure and style of examples closely.

Layout styles

Markdown headings put each part under a heading, which reads well and suits most chat assistants. XML tags wrap each part in tags such as <context>, which Anthropic recommends for Claude and which make it easy for any model to tell instructions apart from pasted material. Plain paragraphs produce a natural message, good for quick tasks.

Worked example

Instead of "write a Facebook post about our new condo launch", the marketing template gives you: a role as a property copywriter, the task, context with the project name, location, price range and key selling points, the audience of first-time buyers aged 25 to 35, constraints to stay under 120 words and avoid unverifiable claims like "best investment", and an output format of a hook line, three short benefits and a call to action. The answer is far more usable on the first try.

Using prompts in apps

If you are building a chatbot or automation, the same structure works as a system prompt. Measure it with the token counter, because a system prompt is sent with every request, and compare revisions with the prompt comparison tool so you know exactly what changed between versions.

Questions