<aside> 🧠
Framework: CONTEXT → GOAL → INPUTS → CONSTRAINTS → OUTPUT FORMAT → REVIEW. Every strong prompt has these six ingredients. Every weak prompt is missing at least two.
</aside>
This short guide covers the working principles behind the prompts in this kit. If you read nothing else, read this once; it will make every prompt in the library more useful.
Almost every prompt in the library follows the same underlying structure. Once you internalise it, you can improve any of these prompts and write your own from scratch.
1. CONTEXT. Who you are and what you are working on. AI models produce generic answers for generic contexts. I am a freelance brand designer working with early-stage SaaS founders is a different starting point from I am a designer.
2. GOAL. What you want the output to be for. Not write a good email but write an email that recovers a stalled proposal without dropping the price.
3. INPUTS. The specific material the AI needs to work with: the client message, the transcript, the numbers, the past behaviour. This is where most prompts succeed or fail; the model can only reason about what you give it.
4. CONSTRAINTS. What the answer must or must not be: length, tone, exclusions, banned phrases, budget, non-negotiables. Constraints usually improve output quality more than adding more information.
5. OUTPUT FORMAT. How you want the answer structured: a ranked list, a table, a message under 100 words, three options at different levels of directness. Vague requests produce sprawling replies.
6. REVIEW. Ask the model to check itself. What are you least confident about? or Which of these claims would change if I removed my largest project? often produces the most useful correction.
The difference between an AI response that reads generic and one that reads like it understands your situation is almost always context volume. Paste the whole message; paste the client history; describe the buyer specifically. The models are not short of processing capacity; they are short of the details only you can provide.
Write me a professional email is a formatting request. Write me a message that acknowledges the delay without apologising, and asks for a decision by Friday is a goal. Goal-oriented prompts produce sharper output every time.
Many prompts in this library explicitly ask the AI to mark whether a claim is supported by the input or is a guess. This is the single most valuable habit to add to prompts you write yourself. It cuts hallucination and helps you know what to trust.
The first response is often not the best one. Follow up with: