Prompt Claro / Clear Prompt Builder

Turn messy thoughts into a first prompt that is clear enough to use, then optionally check the AI answer against the original text.

Prompt Claro, powered by the NHA method, is a small open-source browser tool for people who already use ChatGPT, Gemini, Claude, or any AI tool, but still paste raw, messy thoughts and get inconsistent results.

It helps you turn an initial idea, call note, WhatsApp-style message, client brief, or legal/operations note into a cleaner prompt using a TCE + NHA style workflow.

After the first prompt is generated, it can also help with the next step: paste the AI response back in and create a checking prompt to see whether the AI invented facts, skipped pending information, or drifted away from the original text.

Live demo: https://www.ashishpunj.com/nha-tce/

Why this exists

Most people do not need an "AI agent" on day one.

They first need a better way to ask.

Prompt Claro teaches that step. It gives users a simple pattern:

raw thoughts -> TCE structure -> NHA rules -> better AI output
AI output -> checking prompt -> safer human review

The goal is not to replace judgment. The goal is to make the instruction more explicit, separate facts from assumptions, and force missing information to stay visible.

What it does

• Turns a rough ask, message, or work context into a structured prompt.
• Includes a beginner mode with everyday choices such as replying to someone, studying, improving a resume, planning work, creating content, debugging code, or checking an AI answer.
• Works with ChatGPT, Gemini, Claude, or any AI you use.
• Supports Spanish, English, Hindi, and Punjabi.
• Keeps Spanish examples Mexico-first and English/Hindi/Punjabi examples India-first.
• Generates prompts in the browser.
• Does not require login, install, or API key.
• Includes examples for career/resume, study/postgrad, coding/C++, CX, operations, proposals, legal, and consulting.
• Includes a human checklist before trusting AI output.
• Reveals the answer-checking flow only after the first prompt is ready.
• Generates a checking prompt to audit the AI response against the original text.
• Shows a one-tap star rating after the prompt is copied. The hosted site can record only the rating metadata; prompt text stays local.

Quick start

No setup required.

1. Download or clone this repo.
2. Open index.html in your browser.
3. Write or paste your rough case.
4. Generate the prompt.
5. Copy it into ChatGPT, Gemini, Claude, or any AI you use.
6. Optional: adjust work type, output format, or examples from Options.
7. After the AI answers, paste the response back into the checking box if you want a second-pass review.
8. Generate a checking prompt and use it to review the answer before acting.

Optional check for contributors. This validates JavaScript syntax and runs a small prompt-regression suite for generation, verification, and rating fallback behavior:

npm run check

Privacy

The prompt is generated locally in your browser.

The tool does not intentionally send your case text to a server to generate the prompt, and it does not store that text in a database. The public hosted version may record a simple star rating after use, with language/work-type metadata, but not the text you entered. If you paste sensitive information into any web page or into an AI tool, use your own judgment and review the privacy policy of that AI tool.

Languages

• Spanish
• English
• Hindi
• Punjabi

Language can be selected from the UI. The selected language can also be passed with a query parameter or opened directly through a localized page:

?lang=es
?lang=en
?lang=hi
?lang=pa
/en.html
/hi.html
/pa.html

Project structure

.
|-- index.html
|-- en.html
|-- hi.html
|-- pa.html
|-- info.html
|-- style.css
|-- data/
|   `-- prompt-data.js
|-- prompt-engine.js
|-- app.js
|-- tests/
|-- samples/
|-- docs/
|-- LICENSE
|-- VERSION
`-- CHANGELOG.md

What TCE and NHA mean here

TCE means:

• Task
• Context
• Expectation

NHA means:

• keep the original text visible
• make assumptions explicit
• ask the AI to mark missing information
• keep human verification in the loop

This is intentionally practical, not academic. The project is meant to help people move from casual prompting to a more reliable way of working with AI.

Contributing

Good first contributions:

• improve translations
• add better examples
• add more output formats
• improve mobile UX
• add accessibility fixes
• make the prompt rules clearer
• create classroom exercises around the tool

Read CONTRIBUTING.md before opening a pull request.

Roadmap

See ROADMAP.md.

License

MIT License. See LICENSE.

Copyright (c) 2026 Albert Punj.

Star the project

If this helps you or your students get better AI outputs, please star the repo. It helps more people discover a practical way to move beyond random prompting.