Twelve prompts, a guide and a final-read checklist for writing clear copy with an AI model as the editor: headlines, a brand story, a cold email sequence, a case study, a newsletter and calls to action. Each prompt returns the copy plus a note on what changed and why.
What you build
| Part | What it is | What it does |
|---|---|---|
| Structure | AIDA and problem-agitate-solve | Two classic shapes, short and long |
| Headlines | Headlines and calls to action | Seven headline styles and a set of closing lines |
| Story | Brand story, about page and case study | Your facts, told plainly, with client consent where needed |
| Cold sequence, newsletter and testimonial request | Three emails, one issue with subject lines, three ways to ask | |
| Edit | Rewrite a weak draft, explain a hard idea | Names the three weaknesses and fixes them |
| Finish | A guide and a checklist | Making AI output sound like you, and a last read before sending |
Where this comes from
These are the jobs a one-person business keeps coming back to: write a headline, tell your story, send a cold email and fix a weak draft. The base prompts were written by Hector Diaz. Clara Moretti, his AI assistant, optimized each one so it asks for your real details first, returns a fixed format, and admits what it does not know. Clara is an AI, not a person. Hector reviewed the result.
What you build
Nothing to install. You open a folder, read its README, fill the brackets with your own details and paste the prompt into the assistant you already use. You hand the model your draft or your facts and get back cleaner copy with a short note on what was changed, so the next draft needs less help.
What it deliberately does not do: It does not write from nothing. Every prompt asks for your draft or your facts first.
Do you need this? Be honest first
For someone who writes their own emails, pages and posts and wants them clearer.
Close the tab if you already have prompts you trust for these jobs, or if you want the model to do the thinking without your facts. Every prompt here stops and asks when your input is thin.
What you need
- An AI assistant: Claude, ChatGPT or Gemini. A free tier is enough.
- Your own facts: a draft or the facts behind it, and written consent before you use a client's result or words.
- The time: about 10 minutes per prompt. There are 14 pieces.
What is in the zip
README.md README.es.md the index, in English and Spanish
DISCLAIMER.md DISCLAIMER.es.md as is, no warranty, use at your own risk
01-aida-and-pas/ AIDA and Problem-Agitate-Solve
02-headlines-seven-styles/ Headlines in Seven Styles
03-brand-story-about-page/ Brand Story and About Page
04-blog-post-outline/ Blog Post Outline
05-cold-email-sequence/ Three-Email Cold Sequence
06-client-case-study/ Case Study From a Real Client Result
07-newsletter-issue/ Newsletter Issue and Subject Lines
08-rewrite-weak-draft/ Rewrite a Weak Draft
09-testimonial-request/ Testimonial Request
10-press-release/ Press Release
11-explain-simply/ Explain a Hard Idea Simply
12-calls-to-action/ Calls to Action
13-edit-ai-output/ Editing AI Output So It Sounds Like You
14-final-read-checklist/ Final-Read Checklist
Each folder holds the piece itself in English, plus a README in English and Spanish with a detailed description, the fields to fill in, what comes back, the limits, and a changelog. Every piece is at version 0.0.2.
What was verified and what was not
Checked by script on every piece: each placeholder in a prompt is listed in its README, both languages are present, and none of the marketing words we ban appear.
Not done: the prompts have not been run one by one on Claude, ChatGPT and Gemini and compared. Expect to adjust the wording for your case. The prompts are in English only; add "Answer in Spanish." at the end if you want the reply in Spanish.
Read the prompt before you run it
Read each prompt before you paste it. Ask the model what it plans to return and compare that with the README. A prompt costs a minute to fix. A page of wrong copy costs an afternoon.
