Prompt engineering · ChatGPT & Claude
Stop guessing at prompts. Engineer them.
The PromptCraft Method is a practical guide to the structures, patterns and workflows that make AI output consistent — so the same prompt produces the same quality every time you run it.
$27 one-time · 60-day money-back guarantee
6 Modules
From vague requests to reliable, repeatable prompts.
4 Core Patterns
Templates you can reuse across any task.
5-Step Workflow
A process for engineering a prompt, not guessing.
4-Week Roadmap
Concrete actions to build your own prompt library.
01 / Overview
What is the PromptCraft Method?
Most people prompt AI models the way they would ask a coworker a quick question, and get inconsistent results because of it. The gap between a mediocre answer and exactly the answer you needed is almost always the prompt, not the model. PromptCraft is the set of techniques that closes that gap.
It is built around six modules covering role and constraint design, structured output, few-shot examples, self-critique loops, and how to turn one-off wins into a reusable prompt library. Everything is written to be applied to tasks you already repeat every week — client emails, summaries, code review notes, research briefs — rather than clever demonstrations you will never use again.
“Consistency does not come from a magic phrase. It comes from structure the model can follow.”
Structure over tricks
No magic phrases. You learn the parts of a prompt that actually control output quality.
Built for repetition
Every technique targets prompts you will run again — not one-off experiments.
Model-agnostic
Works the same on ChatGPT and Claude, because it relies on structure, not model quirks.
02 / Audience
Who it is for
Marketers
Produce on-brand copy at volume with prompts that hold tone across every run.
Writers & Editors
Use few-shot examples and critique loops to get drafts worth editing.
Developers
Force structured, parseable output so AI steps can sit inside real workflows.
Founders
Turn recurring operational tasks into saved prompts your team can rerun.
Students & Researchers
Get precise summaries and critiques instead of confident, unusable filler.
AI-curious beginners
Start with a clear mental model of why prompts succeed or fail.
03 / Techniques
Core techniques inside the guide
Role, Context, Constraints
The three-part foundation that makes any prompt predictable.
Detail: name who the model is acting as, what it is working on and for whom, then set explicit limits on length, tone and format.
Structured Output
Specify the exact shape of the answer so it is immediately usable.
Detail: request JSON keys, a markdown table or a fixed heading template, and include one example of the format you expect.
Few-Shot Examples
Show two or three input/output pairs instead of describing the style.
Detail: cover the real range — one easy case and one edge case — and refresh examples when output starts drifting.
Self-Critique Loop
Have the model review its own draft against named criteria, then revise.
Detail: state the criteria explicitly, ask for a rewrite that applies the critique, and run the loop twice for important work.
Uncertainty Handling
Define what the model should do when it does not know something.
Detail: instruct it to mark unknown fields rather than guess, so confidently wrong answers stop looking like accurate ones.
04 / Process
The PromptCraft workflow
- 01
Define the finished output
Write down what a perfect result looks like before you write a single instruction.
- 02
Draft with role, context, constraints
Assemble the three foundation pieces in that order, with concrete limits.
- 03
Pin the format
Add the exact output structure and one worked example of it.
- 04
Run and critique
Generate, then ask the model what is weak or missing against your criteria, and revise.
- 05
Save and version it
Store the working prompt with a task label, and refine it every time it misfires.
05 / Templates
Reusable prompt patterns
The Foundation Pattern
Role + context + constraints as the default opening block for almost any task.
Tip: keep constraints countable — word limits, number of items, banned words.
The Schema Pattern
A declared output schema plus one filled example the model must match exactly.
Tip: add a rule for missing data so gaps are labelled instead of invented.
The Example Ladder
Two to three graded input/output pairs placed before the real request.
Tip: order them easy to hard so the pattern, not the topic, is what transfers.
The Review Pattern
A second turn where the model critiques its own draft, then rewrites it.
Tip: name three criteria maximum — vague critiques produce vague revisions.
06 / Standards
Rules for prompts that hold up
6
Modules of applied technique
4
Reusable prompt patterns
60
Day money-back guarantee
- →Specify the output before you specify the instructions.
- →Replace every subjective word with a measurable constraint.
- →Give one example of the format — never assume it is obvious.
- →Tell the model what to do when information is missing.
- →If a prompt fails, add a constraint instead of rewriting it from scratch.
07 / Application
Your 4-week roadmap
Week 1
Foundations
- ·Rewrite your three most-used prompts with role, context and constraints
- ·Replace every vague adjective with a measurable limit
- ·Log which rewrites changed output quality and why
Week 2
Structure
- ·Add an explicit output format to each prompt
- ·Include one worked example of that format
- ·Add an unknown-value rule to prevent invented details
- ·Test each prompt three times and compare consistency
Week 3
Examples & Critique
- ·Build a two-to-three example set for your hardest task
- ·Run a self-critique loop on your most important output
- ·Apply the critique and save the improved version
Week 4
Your Library
- ·Collect your working prompts in one organised document
- ·Label each with the task it handles
- ·Create variants for your five most frequent tasks
- ·Set a monthly review to refine prompts that drift
Get the guide
Build prompts that work the same way twice.
Full guide access for a one-time payment of $27, delivered by email right after purchase. Covered by a 60-day money-back guarantee — if the method does not improve your output, request a refund within 60 days.
Secure checkout by Digistore24. The debiting is performed by Digistore24.