The 4-Hour Side Hustle · Appendix C

The Prompt Library

Seven working prompts from the book, copy-paste ready. But read the next section first — there's one technique running through four of them that matters more than any individual prompt, and it's the thing you can apply to prompts this book never wrote.

Every prompt in full, with copy buttons. No email needed for any of it.

The one technique worth stealing

Four of the prompts below give the model an explicit way to refuse rather than invent.

WEAK CLIP

The caption prompt. Some clips clear the quality filter and are still fragments that don't say anything complete.

[NEEDS AUTHOR INPUT]

The restructure prompt. Marks a gap in your reasoning instead of filling it.

NO ANSWER IN DOCS

The support prompt. Stops it constructing a plausible answer to a question your documentation doesn't cover.

NEEDS DECISION

Also the support prompt. Anything involving a discount, an exception, or an extension.

A model with no permitted way to say "I can't" will fabricate something plausible, fluently, in your voice, on your account. Giving it an escape hatch costs one line and converts your most dangerous failure mode into a to-do list. Add one to every prompt you write.

When the output is wrong

Five failures, and what actually fixes them. Note that four of the five fixes are about context, not wording.

"It sounds like a press release." Not enough writing samples, or samples that are too formal. Add emails you wrote to a colleague you like — that register is almost always the one your buyers respond to.
"It keeps inventing statistics." Your prompt lacks an explicit prohibition and an escape hatch. Add both. Every prompt above that touches facts has one.
"The output is fine but generic." The context is thin, not the prompt. Generic means it's writing for the average reader of your industry, because that's all it knows. Add raw customer language — the specific, slightly wrong phrasing is what makes output land as yours.
"It ignores half my instructions." Too many rules in one prompt. Split it. The caption prompt works because it does one thing for one platform.
"It was good last month and it's worse now." The project knowledge has drifted — often because your samples now include AI-assisted writing, teaching it to sound like itself. Prune back to your five best human-written samples and re-test.

If output is wrong, add material, not adjectives. Ten pieces of real context beat a hundred words of clever instruction. A sixth writing sample will move the result; a sixth adjective will not — and this is the most common mistake among people who are otherwise good at this.

The clonable Project

Every prompt here assumes a Claude or ChatGPT Project loaded with your Business Brief, five writing samples, and raw customer language. Run them without that and you get competent generic output.

It's a document bundle, so tell me where to send it.

One email a week about building things that run without you. Unsubscribe link at the bottom of every one, and it works.

✓
Sent.

Write the Business Brief before you load anything else. The prompts are the small half — the context is what makes them yours.

© 2026 Jeff Sheikowitz. All rights reserved.