Skip to content
Koit AcademyAll ebooks

AI Tools · PDF · 110 pages

A prompt engineering guide built to outlast whichever model wins next

Twenty chapters on the parts of a request that decide the answer, seventeen of which work on any model you happen to be using.

  • Diagnose a bad answer by which of the four parts was missing, instead of rewording the whole request and hoping.
  • Use constraints as the main lever — the one most people underuse — and get output that lands inside a spec rather than near it.
  • Split a job too big for one prompt into a chain where each step has a verifiable output.
  • Move a working prompt between Claude, ChatGPT and Gemini knowing which three parts need to change and which seventeen do not.
$19.99one-off, no subscription

VAT or sales tax added at checkout where it applies

Get instant access — $19.99

Also in the Koit AI Mastery Bundle — all 5 for $39 (save $47.96).

  • Instant download
  • PDF, works on any device
  • Pay by card or PayPal
  • No subscription
Front cover of Prompt Engineering Masterclass 2026, a 110-page PDF
Look inside ↓

Who this is for

Buy it if

  • You have written plenty of prompts, got inconsistent results, and cannot say why one worked.
  • You use more than one model, or expect to switch, and want the part of the skill that survives the move.
  • You are building something repeatable — a workflow, a Custom GPT, a team standard — and need it to behave the same way twice.
  • You want a method plus a stock of templates, rather than one or the other.

Do not buy it if

  • You want a guide to one product. This is deliberately model-agnostic; seventeen of its twenty chapters apply anywhere.
  • You believe better models will make this obsolete. Chapter 1 argues the opposite and you should read the argument before buying, not after.
  • You only want finished prompts to paste. Buy the 200-template library for seven dollars — it is the same publisher and a fraction of the price.
  • You need image prompting. That is a different craft with its own book; nothing here covers composition, lighting or aspect ratios.

After reading it

What you will be able to do

What is inside

Pages
110
Chapters
20
Parts
2
Prompts
150
Appendices
1
PART ONE — CHAPTERS20
  1. 01

    Why Most Prompts Fail (and the 4-Part Structure That Fixes

  2. 02

    Context Is Everything: Giving the Model What It Needs to

  3. 03

    Role-Setting: Assigning Expertise Without the Roleplay Fluff

  4. 04

    Constraints: The Most Underused Lever in Prompt

  5. 05

    Output Formatting: Controlling Exactly How the Model

  6. 06

    Few-Shot Prompting: Teaching by Example

  7. 07

    Chain-of-Thought: Getting the Model to Actually Think

  8. 08

    System Prompts vs. User Prompts: Where Instructions

  9. 09

    Iterative Prompting: The Refinement Loop

  10. 10

    Prompt Chaining: Breaking Big Tasks Into Steps

  11. 11

    Multi-Step Workflows: Orchestrating Longer Jobs

  12. 12

    Handling Ambiguity: When the Model Should Ask, Not Guess

  13. 13

    Claude-Specific Techniques: XML Tags, Long Context, and

  14. 14

    ChatGPT-Specific Techniques: Custom Instructions, GPTs,

  15. 15

    Gemini-Specific Techniques: Multimodal Prompts and Long-

  16. 16

    Prompting for Writing: Tone, Voice, and Style Control

  17. 17

    Prompting for Code: Specs, Tests, and Debugging Loops

  18. 18

    Prompting for Business: Analysis, Strategy, and Decision

  19. 19

    Prompting for Research: Fact-Finding Without Hallucination

  20. 20

    Building Your Own Prompt Library: Systems That Scale

PART TWO — APPENDIX1
  1. A

    150 Prompt Templates by Use Case

Straight from the book

Two prompts, printed exactly as they appear inside

Template #002 · Follow-up after silencePrompt Engineering Masterclass 2026
Write a follow-up email to a prospect who went quiet after
[last interaction] on [date]. Under 60 words. One clear,
low-friction question. No guilt-tripping language.

Three constraints doing three different jobs: a length cap, a single low-friction ask, and an explicit ban on the tone most follow-ups default to.

Template #017 · FAQ answerPrompt Engineering Masterclass 2026
Write a direct, 2-3 sentence answer to this FAQ: '[question]'.
No hedging, no 'it depends' unless it genuinely depends —
then state exactly what it depends on.

The conditional clause is the craft. Banning "it depends" outright produces confident nonsense; banning it unless it genuinely depends, and then demanding the variable, produces a usable answer.

Sample pages

Three unedited pages from the book.

110

pages in total

20 chapters, all of it written like the three pages beside this one.

Get instant access

The prompt library

150 numbered prompts

Format and delivery

Format
PDF
Pages
110
File size
0.6 MB
Text
Searchable and selectable, not scanned images
Reads on
Phone, tablet, laptop, e-reader — anything that opens a PDF
Printing
Unrestricted
DRM
None. The file is not locked to a device or an account
Delivery
Download link immediately after payment
Access
Permanent — the file is yours once downloaded
Edition
2026
Structure
20 chapters, then Appendix A — 150 templates across writing, coding, business, research and creative

Why $19.99

This is the only book in the catalogue whose value goes up when the tools change, because what it teaches is how to specify work rather than which menu to click. Twenty chapters plus 150 templates for twenty dollars is priced against the alternative of learning the same lessons one disappointing output at a time, across however many models you end up using. The comparison that matters is not against a cheaper book — it is against the hours you will otherwise spend rewriting requests that were never going to work.

One payment, no subscription, and the file is yours to keep. The full delivery spec is above, and the refund policy says exactly when the money comes back.

Refunds

Refunds are handled by Payhip, where the payment is taken. Digital downloads are delivered instantly and cannot be returned, so completing checkout waives the statutory 14-day right of withdrawal. Refunds are still given for a file that will not open, a duplicate charge, the wrong product, or a product that materially differs from its description here — within 14 days, by email.

This wording matches the full refund policy exactly. If the two ever disagree, the policy page is the one that governs — and the difference is what causes chargebacks, so they are kept identical.

Before you buy

Questions about this book

Models keep getting better at understanding vague requests. Isn't prompting a skill with an expiry date?
Better models raised the floor, not the ceiling. A vague request now returns something competent instead of something useless — which is exactly why people stop improving, and why the gap between competent and correct is now the expensive one. That gap is made of context the model cannot know, constraints it cannot guess and a format it has no reason to prefer. No amount of model improvement supplies those, because they are facts about your situation. The chapters likely to age are the three model-specific ones; the other seventeen are about specifying work clearly, which has never expired in any medium.
How is this different from the 200 AI Prompt Templates book?
Method against artefact. That book hands you two hundred finished prompts to fill in and run; it teaches nothing and says so. This one explains why a prompt works so you can build your own for a situation nobody anticipated — and then hands you 150 templates of its own in the appendix. If you only ever buy one, buy this; if you want volume for seven dollars, buy that.
Is it tied to a particular model?
Three chapters of twenty are: one each on Claude, ChatGPT and Gemini specifics. The remaining seventeen are written to work on any of them, and the template appendix marks which model suits a given template rather than assuming one.
Will this help with agents, tool use or the API?
Indirectly at best. Chapters 10 and 11 cover chaining and multi-step orchestration, which is the conceptual groundwork, but there is no code, no function-calling schema and no framework. This is about the words you send, not the pipeline you send them through.
What is the 150-template appendix actually organised by?
Use case, in five groups — writing, coding, business, research and creative — with the templates numbered so you can cite one to a colleague. Both prompts above are from the writing group, printed exactly as they appear.
I have read a lot of free prompting advice. Will this repeat it?
The first two chapters will feel familiar if you have read widely — role, context, task, constraints is common ground. The chapters that usually are not: handling ambiguity so the model asks instead of guessing, deciding where an instruction belongs, and building a library that stays navigable. Chapter 12 in particular is the one most free advice skips entirely.
How long is it, really?
110 pages — twenty chapters and the appendix. Worth saying plainly because it is the shortest of the deep guides and priced the same as the longer ones. You are paying for the method and the templates, not the page count, and if that trade is wrong for you the honest move is to buy the template library instead.
Does it cover prompt injection or security?
No. This is a book about getting good output from a model you are talking to directly, not about hardening a product that exposes one to the public. Nothing here should be read as security guidance.

Read next

Twenty chapters on the parts of a request that decide the answer, seventeen of which work on any model you happen to be using.

Get instant access — $19.99$19.99 · 110-page PDF · instant download