Chapters — Prompt Engineering
Prompt Engineering·Leveling Up·7 min read·Sep 5, 2026

Boundaries and Scope

The five types of constraints that keep an AI response on-topic, correctly formatted, and appropriately scoped.

A boundary is a feature, not a restriction

A prompt with no boundaries doesn't give the model more freedom to help you; it just gives it more room to guess wrong. Naming what you don't want is often what makes a response useful.

Five kinds of constraint
Length & scope
How much you get, and how deep it goes.
Content
What must be included, or explicitly left out.
Style & tone
The voice and formality of the response.
Accuracy & sourcing
Whether it can use outside knowledge, and how it should flag uncertainty.
Behavioral
How it should approach the task itself, e.g. asking questions before answering.
No scope set
Explain how our loyalty points feature works.

A sprawling explanation covering the original design rationale, the points formula, redemption tiers, edge cases, and a history of past changes, most of which the reader didn't ask for.

Scoped to what's needed
Explain how a customer earns and redeems loyalty points, in under 120 words, for a first-time user reading our FAQ. Skip the backend logic: they just need to know how it affects them.

You earn 1 point for every $1 spent, and points show up in your account within 24 hours of purchase. Once you hit 100 points, you can redeem them for a $5 discount at checkout: just toggle "Use Points" before you pay. Points don't expire as long as you make at least one purchase per year.

Content constraint: explain responsibly without over-promising
Explain how compound interest works for someone just opening their first savings account. Don't promise specific returns or make it sound like a guaranteed way to get rich: focus on the mechanism and why starting early matters.

Compound interest means you earn interest not just on what you deposit, but on the interest you've already earned, so your balance grows a little faster each year, not at a fixed rate. It's not a way to get rich quickly; the effect is small in year one and only becomes meaningful over many years, which is why starting early, even with small amounts, matters more than the exact interest rate you get.

Behavioral constraint: ask before advising
Help me reduce customer churn. First ask me 3 clarifying questions about my specific situation. Don't give generic advice until you have my answers.

Happy to help. First, a few questions: 1) What's your current monthly churn rate, and has it been rising or steady? 2) Do customers who leave tend to churn early (first 30 days) or after being active for months? 3) What retention efforts, if any, have you already tried?

Stacking constraints, one layer at a time

  • Layer 1 (basic): "Explain how to price a freelance project."
  • Layer 2 (add length): "...in 150 words."
  • Layer 3 (add audience): "...for someone pricing their first freelance client."
  • Layer 4 (add style): "...practical tone, no jargon, ordered by what to decide first."

Common constraint mistakes

Conflicting constraints
"Comprehensive analysis in under 30 words": pick one; they can't both be true.
Over-constraining
Ten rules stacked on a simple task usually produces a worse answer, not a better one. Keep to the 3–4 that matter.
Under-constraining sensitive topics
Medical, legal, or financial topics need explicit tone and disclaimer guidance, not just a scope.
Vague constraints
"Keep it professional" means something different to everyone. Say what professional means for this response.
Part of a free guide

Prompt Engineering

A simple guide to writing better prompts, so AI gives you the answer you actually want.

Browse All Guides →