For anyone using a free or cheap AI model and wondering why the answers still sound like guesses

The Model Is Rented. The Playbook Is Yours.

19 copy-paste prompts that install the same reasoning standard into any model, Claude or ChatGPT. Each one comes with a filled example and what good output looks like, so you know the moment it's working.

$7 · one-time

You've read the free playbook. Now you have to type it out every time.

The Fable 5 Playbook taught you the four-part method: verify before you answer, plan before you execute, contract before you delegate, format before you hand it back. That part is free and it's real. But knowing the method and having the exact words ready in the middle of a task are two different things. Most people read the method once, nod, and then go back to typing plain questions into a cheap model and getting a plausible paragraph back instead of a checked answer.

  • You paste a question into a free model, get a confident answer, and only find out it was wrong after you've already acted on it.
  • You know you're supposed to ask it to verify its claims first, but rewriting that instruction from memory every session gets skipped by the third question.
  • You're one subscription upgrade away from thinking the problem is the model, when the problem is that nothing you're typing asks it to check itself.
  • The four-part method sits in your notes somewhere. It doesn't install itself.

The method, turned into 19 prompts you paste instead of write.

Each prompt installs one piece of the four-layer method into whatever model you're using: reasoning that verifies its load-bearing claims before it answers, workflows that plan then check each step instead of running blind, prompt contracts that give sub-tasks and helpers a tight scope so they don't wander, and deliverables that answer first, rank the findings, and label every number CONFIRMED or ESTIMATE. Every prompt comes with a filled example of what it catches and what good output looks like, so you're not guessing whether it worked.

What's actually inside

The one-paste master install

Paste this at the top of any new chat or agent session and it installs all four layers at once: verify claims, plan then check, contract sub-work tightly, and format the final answer. No more re-explaining your standards three prompts in.

Verify before you build

Forces the model to name the 2 to 3 claims an answer actually depends on and check those specifically before it writes a word, instead of answering first and hoping the details hold up.

The trust-order enforcer

Makes the model tag every claim TOOL, FILE, or MEMORY, so you can see what it actually just checked versus what it's recalling from training and passing off as current.

Find the mechanism

For the moment a result surprises you, a report that shows zero, a number that vanished. Names the 3 most likely specific causes, ranked, each with a two-minute check, instead of a vague guess.

The 95 gate

Before you trust a conclusion enough to act on it, this makes the model score its own confidence out of 100. Under 95, it has to say HOLD and name the exact evidence still missing instead of rounding up to sound done.

The deliverable formatter

Turns any answer into: verdict first, ranked findings, every number labeled CONFIRMED or ESTIMATE, one next action. No 'great question,' no wall of text you have to dig through.

The sparring partner

Blocks the model from agreeing with your idea by default. It has to build the strongest objection first, then give you one of three honest responses, including a sharper version of your own plan when one exists.

The blind-spot scan

Run before you finalize anything you're already too close to see clearly. Names the one or two things you're not seeing because you're inside the problem, and the earliest point you could have caught it.

This is for you if

  • You already use Claude or ChatGPT most days and you're tired of getting confident answers instead of checked ones.
  • You went through the free Fable 5 Playbook and want the method as ready-to-paste prompts instead of rewriting it from memory every session.
  • You'd rather spend $7 fixing how you prompt than pay more each month hoping a pricier model reasons better on its own.
  • You want a filled example for each prompt so you know what it's supposed to catch before you rely on it.

Not for you if

  • You haven't read the free Fable 5 Playbook yet. Start there, it's free, and this system is that method turned into prompts, not a replacement for it.
  • You want a tool that runs itself. Every prompt here still requires you to paste it and read the output.
  • You're looking for prompts that make a model sound more impressive. This system makes it more accurate, which sometimes means slower, more honest answers.

Get The Fable-Grade Prompt System

  • 19 copy-paste prompts, each with a filled example and what good output looks like37 cents a prompt
  • The one-paste master install for all four layers: reasoning, workflows, prompt contracts, deliverablesinstall once, reuse in every session
  • The Master Start Here prompt, which reads your role and picks the 3 prompts that matter most for youno guessing where to begin
  • Works in Claude, ChatGPT, Gemini, or any chat or agent model you already haveno upgrade or new subscription needed

$7

If it's not useful, email us

Run the master install in your next real conversation. If you don't think it changed the answer you got, email us within 14 days and we'll refund the $7. No forms, no reasons needed.

Questions

How is this different from the free Fable 5 Playbook?

The free playbook teaches the four-part method: verify, plan and check, contract sub-work, format the deliverable. This system is that same method already turned into 19 exact prompts, each with a filled example and what good output looks like. The free version teaches you the method. This hands you the 19 prompts already written, ready to paste.

Does this work with ChatGPT, or only Claude?

Both, plus Gemini and any other chat model. Every prompt is written in plain language a model reads and follows, not a Claude-specific feature. A few of the 19 assume you're working in an agent or coding tool like Claude Code or Cursor, and those come with a chat fallback so you can still use them in a plain chat window.

I'm not technical. Can I still use these?

Yes. Most of the 19 prompts are copy, paste, and fill in your own details in brackets. No code, no setup. The ones that assume an agent harness are marked, and each of those includes what to do instead in a plain chat.

Do I need to use all 19 every time?

No. The Master Start Here prompt, the one to run first, asks about your role and your most expensive recurring task, then tells you the 3 prompts that matter most for you and the order to run them in. You don't have to memorize all 19 to get value from the first one.

What's the actual difference in the answers I get?

Without these, most models answer immediately and sound confident whether they checked anything or not. With the verification and 95-gate prompts installed, the model has to name what it actually verified, label estimates as estimates, and say HOLD instead of guessing when it's not sure. The example inside each prompt shows exactly what that looks like.

Is $7 really it?

Yes. The system is $7, one time, delivered by email. Nothing about it auto-renews or bills again.