Part 2 of 5

Prompting Moves

Three small habits that change the quality of an answer more than any clever wording. Each one shown before and after — copy the "after" and make it your own.

Tell it about you

An AI knows an enormous amount about the world and nothing about your branch, your budget, or your community — until you say so. Context is the cheapest upgrade there is.

Before — generic
Plan a summer reading program.
After — yours
Plan a summer reading program for our branch.

We serve a working-class neighborhood with many
multilingual families. Budget is about $1,500.
Most attendees are kids 5–12 with younger siblings
in tow. Last year's theme was "Oceans" and our
best-attended event was a touch-tank visit.

Give me a theme, four weekly events, and one idea
to reach families who've never come before.

Ask in Markdown

Structure in, structure out. Headings and bullets tell the model exactly what you want and in what shape — and they're easier for you to edit and reuse, too. More on why in Part 4.

Before — run-on
Write a flyer for our author talk with the date and
time and where it is and that it's free and people
should register and also mention parking and that
there will be a book signing after and refreshments.
After — structured
Write a flyer for our author talk.

## Details
- Author: [name]
- Date & time: [when]
- Location: [room]
- Cost: Free — registration required

## Must mention
- Free parking in the rear lot
- Book signing & refreshments after

## Tone
Warm and inviting, about 80 words.

Let the agent write it

You don't have to craft the perfect prompt. Ask the model what it needs, or have it write the prompt for you — then answer its questions. It's the reference interview, run in reverse.

Before — guessing alone
Write our annual report.

(…then re-writing the prompt five times,
trying to think of everything it needs.)
After — let it lead
I need to write our library's annual report.

Before you draft anything, ask me up to 7 questions
that would help you do this well — about our
audience, what we accomplished, the numbers I have,
length, and tone. Wait for my answers, then we'll
write it together.

This is sometimes called meta-prompting — prompting the model about how to prompt. When you're stuck, it's the fastest way forward.