Prompting Fable 5

Six habits for prompting Claude Fable 5 — give it the why, negative prompting, letting it act, proof-of-done, the reasoning trap, and saying less — with paste-ready prompt blocks.

Six habits for getting the most out of Claude Fable 5, distilled from Anthropic’s prompting guide by Nate Herk in How Anthropic Engineers Actually Prompt Fable 5. Five of them work on any model; one is Fable-specific. Each comes with a paste-ready block — the long-term move is to bake these into CLAUDE.md, your skills, and your agents rather than retyping them per prompt.

The premise: Fable follows short, clear direction better than older models, so the leverage has moved from spelling everything out to pointing it at the right intent.

1. Give it the why — any model

One line of “here is why I am asking” does more than piling on step-by-step detail. Fable does better when it understands your intent — context lets it connect the task to the right information instead of guessing what you meant.

Instead of “Write me an email to a client about the delay.”:

I'm working on [the bigger task] for [who it's for]. They need [what the
output enables]. With that in mind: [your request].

2. Tell it what NOT to do — any model

A more capable model acts on its own more readily: drafting emails, making backups, refactoring around a one-line fix. The sentence that saves you starts with “do not”. Anthropic’s own guide is full of negative prompting — “don’t add features”, “do the simplest thing that works”.

Instead of “Take a look at this problem and handle it.”:

When I'm describing a problem or asking a question, the deliverable is your
assessment. Report what you find and stop. Don't fix, send, edit, or delete
anything until I say go. Do the simplest thing that works, and skip cleanup
I didn't ask for.

3. Let it act once it has enough — any model

More deliberation is not better. On a model that can run for minutes, endless option-surveying just burns time and money on choices it will never use.

Instead of “Research everything and make a full plan before you do anything.”:

When you have enough information to act, act. Don't re-derive what we've
already settled or narrate options you won't pursue. If you're weighing a
choice, give a recommendation, not an exhaustive survey.

Match the effort level to the task

Fable’s effort settings are part of the same habit — don’t pay for deliberation the task doesn’t need:

The same logic applies one level up: you probably only need to reach for Fable at all for maybe 5–15% of tasks. Using it for everything is overkill, especially once you’re paying usage credits.

4. Make it prove it — any model

“Done and working” is a claim, not a fact. Anthropic’s own testing found this one line nearly eliminated made-up status reports, even on tasks built to provoke them. This is the one most worth baking into every skill and agent you own.

Instead of “Is it done and working? Great, thanks.”:

Before you tell me something is done, point to the result that proves it.
Only report work you can show evidence for. If something isn't verified,
say so plainly instead of guessing.

5. Stop asking it to “show its reasoning” — Fable-specific

This one is about a line to remove. The video reports that a standing “explain your reasoning” instruction — especially in a system prompt — can trigger a refusal or get the request silently rerouted to Opus 4.8: Fable runs a safety check before answering, and asking the model to reveal its private reasoning is one of the buckets that trips it (alongside anything that looks like hacking or dangerous biology). On the API the response shows which model actually answered; in the apps you may not notice the swap.

There’s no paste block here — just audit your CLAUDE.md and system prompts for standing “show/explain your reasoning” lines and take them out.

6. Say less, not more — any model

One line beats twelve. Because Fable is smart enough — especially wrapped in a good environment of context, tools, and skills — a short instruction now steers as well as spelling out every rule by name. This doesn’t contradict habit 1: adding the why is one sentence of context, not bloat.

Instead of “Rule one: be concise. Rule two: … Rule three: …”:

Lead with the outcome, keep it simple, and pause only when the work truly
needs me.