/08

Prompting

A prompt builder based on the same things a capable person needs: job, context, material, output, and quality bar.

Start here

7 min

Open +

Prompt engineering sounds technical. Most of it is simply briefing clearly, providing the missing context, and defining what good looks like.

Leave knowing / Write prompts that survive real work

The five parts

A strong prompt answers five normal questions.

Job

What exact thing should be done?

Context

What does the AI need to understand about the situation or audience?

Material

What source information should it use and what should it ignore?

Output

What should the answer look like: length, format, fields, structure?

Quality bar

What makes the result good, and what must it never invent?

When the answer is bad

Fix the brief before blaming the model.

Did I ask for one clear job or five mixed jobs?

Did I provide the facts the answer depends on?

Did I define the audience and purpose?

Did I show the structure or example I want?

Did I tell it what to do when information is missing?

Am I asking a language model to solve a problem that should be a formula, database query, or normal software rule?

Try it

JobContextSource materialOutputQuality bar

Your brief

JOB
Turn these meeting notes into a useful follow-up

CONTEXT
The audience is a busy client who needs decisions and next actions, not a transcript

SOURCE MATERIAL
Use only the notes I paste below

OUTPUT
Five bullet summary, decisions, actions with owners, open questions

QUALITY BAR
Do not invent facts or owners. Mark missing information as Needs confirmation.

If the source does not support an answer, say what is missing instead of guessing.

Notes / Subscribe

No schedule. No content calendar. A note goes out only when there is something worth understanding.