Stop hoarding prompts.
Learn the four moves.
The Prompt Pattern Playbook: four named moves to fix what your AI gets wrong.
You have a folder of saved prompts, and the AI still hands you work you cannot trust: a status that reads "on track" over a contradiction, an owner it invented from a job title, a confident sentence that was a guess in disguise. This book teaches four reusable moves that fix exactly those failures, so you stop copying prompts and start knowing which move the moment calls for.
A copied prompt works once. A move works everywhere.
A prompt you saved is welded to the task it was born for. The moment the task shifts, you are stuck, because you never learned the move underneath. Patterns travel with you. This book teaches four of them, each one a named move that fixes a specific way AI output goes wrong, so by the last page you can look at a bad answer, name what failed, and write the prompt that fixes it without the book open.
Four failures. Four moves. One for each.
Every move is taught the same way: the failure it fixes, how it works, a real before-and-after, when not to use it, and a drill. Here is what each one is for.
The AI fills gaps by default. It resolves a missing owner or an unconfirmed status with something plausible, in the same calm voice it uses for real facts, so the fabrication hides in plain sight.
The move: give the model explicit permission to be incomplete, so it writes UNKNOWN where the source is silent instead of papering over the hole.
A guess and a fact arrive in the same certain voice, so partial evidence drifts up to sounding like proof and you act on tone instead of evidence.
The move: make the AI grade every claim HIGH, MEDIUM, LOW, or UNKNOWN, tied to a source you can check. The line that does the work: never HIGH without a quotable source.
The model is trained to be agreeable, so it validates whatever you hand it and goes useless at the exact moment you wanted it to find the holes, before you commit.
The move: tell it not to be agreeable, give it a structure that forces specificity, and give it an honest out, so a genuine pass actually means something.
Asked to help prepare a call, the AI reaches past the draft and makes it, taking ownership of a decision that should stay with you.
The move: have it prep options, trade-offs, and evidence, then stop at a blank DECIDE: [ ] line and hand you the pen.
Every move comes with a worked before-and-after.
No move is asserted without showing the weak prompt, the patterned prompt, and the difference in the output. Here is CONFIDENCE on three lines that all sounded equally sure.
Before: ungraded
The launch is on track. Sarah owns the integration. We are within budget. (three claims, one confident voice, nothing to check)
After: the CONFIDENCE move
Within budget = HIGH
("tracking 4% under", May summary)
Sarah owns it = MEDIUM
(named in kickoff, unconfirmed)
Launch on track = LOW
(Apr 30 standup: "slipping ~2 weeks")
The line that sounded most certain had the least behind it. You can see it now.
The moves are the start. The system is the book.
Five recipes for stacking the moves in the right order. Ground, then grade. Pressure-test, then hand over the pen.
Four questions that map any bad AI output to the one move that fixes it. The skill that matters most is knowing which to reach for.
One move a day, then combinations, then a self-test, then real work with the book closed. Reading becomes reflex.
A light bridge from a single prompt to the standing instructions inside systems and agents. The same four moves scale up.
A book you read once, then a toolkit you keep.
Instant download. Read it on anything.
The one-line definition of every move is free on nerdychefs.ai, and Noodle, our prompt coach, will hand you examples for nothing. This book is the part the one-liner cannot fit: why each failure happens, a real before-and-after, the when-not-to logic, the combinations, the decision tree, and the drills that turn reading into reflex.
Before you buy
Is this a pile of prompts to copy?
Do I need the free site or Noodle to use it?
What formats do I get?
Will it age out when the models change?
Who is it for?
Close the folder. You will not need it much longer.
Learn the four moves, and a bad AI output stops being a mystery. You name the failure, reach for the fix, and trust what you ship.