foundations · Level 2
Prompt like the highest-paid engineer in the room
The eight-part structure elite prompt engineers use, worked through five real outcomes from weak prompt to strong: a website, an app, an investor document, an image or video, and a difficult email.
Start with the essentials
The short answer
An elite prompt is not a longer prompt. It states what good looks like, what bad looks like, and how you and the model will both check the result, before a single word of output exists. This workbook teaches that structure through five weak-to-strong examples: a designed website, a working app, an investor document, directed image and video, and a difficult email that moves a negotiation forward.
What you will learn
- You will be able to name and use the eight parts of an elite prompt: role, context, constraints, output format, examples, iteration, verification and failure modes.
- You will be able to brief an AI coding assistant for a website section by section, using a small design system, so the result reads as designed rather than templated.
- You will be able to prompt for a working app or tool with acceptance criteria you wrote yourself, and name the two most common AI code failure modes before you hit them.
- You will be able to write an investor or board document that contains no invented numbers, by fact-locking the prompt and asking the model to flag gaps instead of guessing.
- You will be able to direct AI image and video generation with camera, lighting and composition language, and check the results against known failure points.
- You will be able to prompt for a difficult email or negotiation by stating your real objective, running a pre-mortem on the other side's objections, and comparing tone options.
Who it is for
Anyone who already gets useful answers from AI and wants outcomes that read as expert-crafted rather than AI-generic: people briefing an AI coding assistant on design work, builders prompting for working tools, founders and managers drafting material for investors or a board, anyone directing AI image or video generation, and anyone who needs a difficult message to actually land. Comfort with basic prompting is assumed.
Before you start
- Comfortable writing a basic prompt using role, task, context, format and constraints, as covered in Your first prompts. No coding background is required. Chapter two assumes you can open and run a single HTML file in a browser, as in Build your first app with an AI coding assistant, but you can also read that chapter for the prompting pattern alone.
Keep learning
The complete workbook
This workbook takes prompting from a beginner habit to a discipline. You will learn the eight parts behind consistently excellent AI output (role, context, constraints, output format, examples, iteration, verification and failure modes), then apply them across five outcomes that matter at work: briefing a premium-looking website section by section, prompting for a working app or tool, writing an investor or board document without a single invented number, directing image and video like a photographer rather than describing a subject, and writing a difficult email that pre-empts the other side's objections. Every outcome is shown as a weak prompt, why it fails, and the rebuilt strong prompt, so you learn the diagnostic skill rather than a template to copy.
- 01The eight-part structure behind every elite promptIn the workbook · 1 exercise
Anyone can type a request into an AI model. The people who get expert-grade results, every time, are not luckier with wording. They are working to a structure, and they check their own output the way a good editor checks a junior colleague's draft. This chapter sets out that structure. The next five chapters put it to work on five real jobs: a website that does not look like a template, a working app, a document an investor would actually read, an image or video that looks directed rather than generated, and an email that moves a difficult conversation forward.
- 02Prompting for a website that reads premium, not template-genericIn the workbook · 1 exercise
Every AI model has seen thousands of software-product landing pages and near-identical templates. Ask for 'a modern, professional website' and it hands you the average of all of them: a centred hero, a gradient background, three icon boxes in the same rounded style, and a testimonial carousel. None of that is exactly wrong. It is just instantly recognisable as generated. This chapter shows you how to prompt for something that reads as designed, and how to brief an AI coding assistant so the result holds together section by section instead of drifting.
- 03Prompting for a working app or tool, not a demo that looks like oneIn the workbook · 1 exercise
Build your first app with an AI coding assistant taught you to brief, build in small steps, review and test. This chapter is about the prompting discipline that makes that whole process reliable at a senior level: a hard scope boundary, acceptance criteria you write before any code exists, and two AI coding failure modes worth naming out loud rather than discovering the hard way.
- 04Prompting for a document an investor or a board would actually readIn the workbook · 1 exercise
Business writing has a failure mode the other four outcomes in this workbook do not share as sharply: a model asked to sound 'compelling' about a company it barely knows will invent specifics to sound convincing. A market-size figure, a growth rate, a competitor claim, stated with exactly the same fluent confidence as a real one. This chapter is built around removing that risk entirely, not just writing better sentences.
- 05Prompting for images and video that look directed, not generatedIn the workbook · 1 exercise
Ask an image model for 'a professional image of a business meeting' and you get the single most common arrangement in its training data for that phrase: a symmetrical group, evenly lit, everyone facing camera. Recognisably generated, at a glance. This chapter replaces vague description with the vocabulary a photographer or a director actually works in, plus a disciplined way of refining what comes back.
- 06Prompting for a difficult email or a negotiation that actually movesIn the workbook · 1 exercise
Write a firm email to a client who hasn't paid and you will get exactly that: firm, and generic. It will not tell you whether firm is even the right register for this specific person, and it cannot pre-empt an objection it was never told about. This chapter treats a difficult message as a small negotiation with a stated objective, not a tone to request.
- 07What you can do nowIn the workbook · Reading
You have five weak-to-strong examples behind you now, plus the eight-part structure that produced every one of them. Here is what to do with that structure next.
Also inside: a 10-point checklist, a glossary of 12 terms and 12 questions and answers to test yourself. 6 hands-on exercises, each with a worked answer at the back where the workbook gives one.
No login, no card, no account. Before the download we ask you to follow Mickai (two quick links). Free to download and use for personal learning, study groups and inside your own team. Please do not resell the workbooks or republish them as your own. Link people to trust-agent.ai instead.
Test yourself
Questions and answers
What are the eight parts of an elite prompt?
Role, context, constraints, output format, examples, iteration, verification and failure modes. The first six build on basic prompting; iteration, verification and failure modes are what push a result from competent to expert-grade.
What does naming a failure mode inside a prompt actually do?
It tells the model, in advance, the specific way this exact kind of task tends to go wrong, such as a generic centred layout or an invented number, so it can actively avoid that pattern rather than you catching it after the fact.
Why does asking for a 'modern, professional' website tend to produce a generic template?
Those words carry no layout, type, spacing, colour or motion meaning on their own, so the model falls back on the most common pattern in its training data for that phrase: a centred hero, a gradient background and rounded icon boxes.
Why build a page one section at a time instead of asking for the whole thing in one prompt?
A single huge prompt asks the model to hold every section consistent across a much longer generation, with no chance to catch drift until the end. Staged briefing catches mistakes in minutes and lets you restart one section without losing the rest.
What is excessive agency, and why does it matter when prompting an agent that can run commands?
It is a named risk where an agent, given a loose instruction and real permissions, takes actions well beyond what was intended, including changing or deleting files outside the scope you meant. The fix is a narrow scope, a plan requested before any code, and never handing an agent broader access than the job needs.
Why should acceptance criteria be written by you before the prompt, rather than asked of the AI afterwards?
If you cannot state what a correct result looks like for a given input in advance, you have no way to tell a right answer from a merely fluent-looking one once the assistant produces something.
What is the most dangerous failure mode in AI-written business or investor documents?
Invented precision: a model asked to sound compelling about a company it barely knows will fabricate specific figures, market claims or competitor details, stated with exactly the same fluent confidence as a real one.
What should a model do when a claim you want in the document is not in the facts you supplied?
Flag it explicitly, for example by writing [NEED: description of what is missing], rather than estimating or rounding to something that sounds plausible.
Why does a vague image prompt tend to produce a recognisably AI-generated look?
Without composition, lighting or camera instructions, the model defaults to the single most common arrangement in its training data for that subject, which is exactly what makes it instantly recognisable as generated.
What should you check for deliberately when reviewing a generated image, rather than just glancing at it?
The parts most likely to be wrong and most likely to catch a viewer's eye first: hands, any visible text, reflections and shadows for consistency with the stated light source, and repeated or tiled background patterns.
Why is 'write a firm email' not enough to get a message that actually achieves your objective?
Firmness is a tone, not an outcome. Without a stated objective, the model cannot tell whether you need full payment by a date, a payment plan, or simply a documented request, and will default to a generic tone that may not fit any of them.
What is a pre-mortem in this context, and what does asking for one before the draft actually do?
Asking the model to list the other side's likely reasons and reactions before it writes anything. It surfaces objections you might not have thought through yourself, so the message can pre-empt at least one of them instead of triggering it.
When you have finished
Get your certificate of completion
Type your name and download a certificate for this workbook as a PDF, ready to print or to add to LinkedIn. It is made on your own device, so your name is never sent to us. It is a self-declared certificate, not an accredited qualification.
Learn the language
Key terms
- Role prompting
- Telling the model the specific expertise and standard to write to, such as a named level of seniority, rather than a generic persona.
- Context
- The real background facts the model cannot infer on its own, including the constraints of the actual situation, not just who the answer is for.
- Constraint
- A limit in a prompt that states both what to include and, just as importantly, what to actively avoid.
- Output format
- The exact shape an answer should take, including how a larger piece of work should be delivered in stages rather than all at once.
- Few-shot example
- A worked sample included in a prompt so the model matches a demonstrated standard rather than a described one.
- Iteration loop
- A refinement plan decided before you start: how many rounds, what changes between them, and how you will choose the best result.
6 of the workbook's 12 terms. The complete glossary is in the workbook.
Follow the evidence
Sources and checks
Facts last checked: .
These workbooks use AI assistance. See how the workbooks are made.
- LLM06:2025 Excessive AgencyOWASP Gen AI Security Project
- LLM09:2025 MisinformationOWASP Gen AI Security Project
- The 'vibe coding spectrum' approach to AI-assisted software developmentNational Cyber Security Centre (NCSC)
Created by Mickarle Wagstaff-Irons - Micky Irons with the Mickai team. Published by Mickai LTD. Last updated 29 September 2026.
NextKeep going
Where to go next
Recommended for you
Build and launch a website with AI
Plan a simple site, have AI write the HTML and CSS, understand and check the code, then publish it free with HTTPS and sound search, accessibility and privacy basics.
Recommended for you
Build your first app with an AI coding assistant
Build a small bill-splitting web page with any AI coding assistant while staying in control: brief it, build in steps, review the code, test edge cases and keep safe copies.
Recommended for you
AI images and video: rights, provenance, labels and quality checks
Use AI images and video responsibly: ownership in the UK, EU and US, tool terms, consent, labelling, Content Credentials, quality checks and simple records.