Skip to content
47% off: the complete course for $79 (reg. $149)Order now →
Agent Builder
Module 3Full course10 min read

A Library of Prompt Patterns That Work

Battle-tested prompt patterns you can reuse — interview-me, critique-then-revise, explain-your-assumptions, and more — with when to use each.

Most people who are good with AI aren't writing brilliant new prompts every time. They're reaching for a small set of moves that reliably work: make the model ask before it guesses, make it show its assumptions, make it criticize its own draft before you see it. Experienced practitioners use a dozen of these moves so often they stop noticing them.

This lesson writes those moves down. You'll get a library of named prompt patterns, each with a copyable prompt, an explanation of why it works, and clear guidance on when to use it and when it's a waste of time. The names matter: once you can say "this needs a pre-mortem" or "run critique-then-revise," you'll pick the right tool in seconds instead of rewriting prompts from scratch.

How to use this library

A prompt pattern is a reusable shape for a request that solves a recurring problem, the way a recipe solves "what's for dinner." Each pattern below is written as a fill-in template: square brackets mark what you replace. Patterns are grouped by the job they do:

JobPatterns
Clarify the task before work startsInterview me, Explain your assumptions, Restate before you start
Think more carefully about the problemPlan first, Options before answers, Pre-mortem
Ground the answer in real materialQuote then answer, Permission to say "I don't know"
Improve a draft before you see itCritique then revise, Grade against a rubric, Simulated reader
Reuse what workedWrite the prompt for me, Save the recipe

Two rules apply to every pattern. First, patterns sit on top of good context; they don't replace it. If the model is missing the facts, no pattern will save you (see From Prompt Engineering to Context Engineering). Second, every pattern costs something: an extra round trip, more output, or more of your attention. Use them where the stakes or the ambiguity justify it, not by reflex.