Skip to content
FORGENO ADS GAMESUnlock all $8

AI Skills Academy · Unit 3 of 8

Unit 3: Writing great prompts

The five parts of a strong prompt, examples, step-by-step requests and editing like a pro.

Opens the first lesson of this unit in AI Skills Academy. Every lesson is open from the start.

Lesson 1: The five parts of a strong prompt

Goal: Build a prompt with a task, context, format, constraints and examples.

A prompt is just a request. But the model only knows what you tell it, so vague requests get generic answers. Strong prompts usually have up to five parts:

PartThe question it answersExample
TaskWhat exactly should it do?"Write a flyer..."
ContextWho is it for and why? What background matters?"...for my dog-walking service; readers are busy parents in my neighborhood"
FormatWhat shape should the answer take?"A headline, three bullet points and one closing line"
ConstraintsLimits and must-haves"Under 120 words, friendly tone, include my price of $12 per 30-minute walk"
ExamplesWhat does good look like?"Here is a flyer I like the style of: ..."

Worked example: weak to strong

  1. Weak: Write about my dog walking business.
  2. Strong: Write a flyer for my new dog-walking service. Readers are busy parents in my neighborhood. Format: a catchy headline, 3 short bullet points and a closing line telling them to text me. Keep it under 120 words and friendly. Include: $12 per 30-minute walk, weekdays 3 to 6 pm, and that I have walked my neighbors' dogs for two years.

The strong prompt is longer, but every line removes a guess the AI would otherwise make.

Hands-on lab in the room: Prompt builder. Build a prompt piece by piece. The checker shows which parts are present and how specific they are.

Not every prompt needs all five parts. A quick question needs only a clear task. But when the answer matters, run through the five and ask: "What would a smart stranger need to know to do this well?"

Remember: Task, context, format, constraints, examples. Every missing part is a guess the AI has to make.

Lesson 2: Show, do not just tell: examples

Goal: Use one to three examples to get a consistent pattern.

Sometimes describing what you want is harder than showing it. When you include a few examples of input and the output you want, the model copies the pattern. This is called few-shot prompting: "few shots" meaning a few examples.

Worked example: sorting customer messages

Sort each customer message as PRAISE, PROBLEM or QUESTION.

Message: "The delivery was two days late." -> PROBLEM
Message: "Do you deliver on Sundays?" -> QUESTION
Message: "Best pizza in town, thank you!" -> PRAISE

Message: "My order was missing the garlic bread."
->

The model answers PROBLEM, in exactly the same format, because the examples set the pattern.

Examples work for style too. Paste two posts you wrote and say "write a new post in the same style about our Saturday sale." That beats trying to describe your style in adjectives.

Watch out: Models copy examples closely, including their mistakes and their quirks. If all your examples are about pizza, the answers may drift toward pizza. Use examples that differ from each other in the ways that do not matter.

Few-shot prompting
Including a few examples of input and desired output in a prompt so the model follows the same pattern.
Zero-shot
Asking with no examples at all, just instructions.

Remember: One to three varied examples beat a paragraph of description.

Lesson 3: Break big jobs into steps

Goal: Split complex work into a chain of smaller prompts.

A model does better when a big job is broken into steps. There are two ways to do it.

Ask it to work in steps. For math, logic or planning, ask it to work through the problem step by step before giving the final answer. Writing out the steps tends to improve accuracy, and it lets you see where it went wrong.

Chain your prompts. Do the job as a series of prompts, each feeding the next: outline, then draft, then improve. You check each step before moving on, so a mistake early on does not spoil everything.

Worked example: a cover letter, in four prompts

  1. Paste the job post: "List the five most important requirements in this job post."
  2. "Here is my experience: [your real experience]. Match each requirement to something I have done. Tell me where I have no match."
  3. "Write a 200-word cover letter using only these matches. Do not invent experience."
  4. "Make it sound more like a confident 19-year-old and cut any cliches."

Each step is easy to check, and the final letter is built only on true facts about you.

Notice step 3 says "do not invent experience". Constraints like that matter most when the AI is writing about you.

Remember: Big job? Outline, check, draft, check, improve. Small steps are easier to verify.

Lesson 4: Edit like a pro

Goal: Improve an AI answer with specific feedback, questions and roles.

Treat the first answer as a first draft. Professionals rarely use it as is; they push it closer with specific feedback.

Vague feedbackSpecific feedback
"Make it better.""Cut it to 3 sentences and lead with the price."
"Too formal.""Write it like a friendly text to a neighbor. No words like 'regarding' or 'kindly'."
"I don't like it.""Keep bullet points 2 and 3. Rewrite point 1 so it mentions weekends."

Three more moves that work well:

  • Let it ask you first: "Before you start, ask me up to 5 questions you need answered to do this well." It will often ask things you did not think to say.
  • Ask for options: "Give me 3 different openings: one funny, one serious, one short."
  • Give it a role for a different point of view: "Act as a hiring manager reading this resume. What would make you reject it?" A role changes the angle. It does not give the AI real expertise or experience, so still check the advice.

And know when to stop: if it keeps missing after a few rounds, the problem is usually missing facts. Add them, or write that part yourself.

Remember: First answer = first draft. Specific feedback gets specific improvements.

Practise it in AI Skills Academy

Each lesson ends with two or three questions that check you can use the idea, not just recognise it. Wrong answers explain why and come back in Review. The unit ends with a 5-question quiz; pass it with 80% or more to master the unit. Hands-on lab: Prompt builder.

Every lesson is free: no ads, no lives and no paywall.