Prompt Engineering Guide

Prompting is one of the most important skills when working with AI. A well-written prompt helps the AI understand what you’re building, how you want it built, and what should not change. Better prompts generally lead to:

Higher quality outputs
Fewer retries
Lower credit consumption
Faster development
More consistent results

How AI Understands Your Prompt

Every prompt goes through a reasoning pipeline before the final output.

Think of AI like a developer joining your project. The more information you provide, the better it can understand your intent. The AI works best when your prompt includes:

Context

Explain what you’re building.

Goal

Describe what you want to achieve.

Requirements

List required features.

Constraints

Explain what should NOT change.

Examples

Reference previous screens or designs.

Output

Specify the desired format or result.

The Anatomy of a Great Prompt

A good prompt usually follows this structure:

Context

Goal

Requirements

Constraints

Reference

Expected Output

Example: Poor Prompt

Make my website better.

Problems:

No context
No requirements
No design direction
AI must guess

Result:

Generic output
More retries
More credits

Example: Great Prompt

I'm building a SaaS landing page for a CRM platform.

Update only the homepage.

Requirements:

- Improve the hero section
- Add customer testimonials
- Keep existing navigation
- Preserve branding colors
- Do not modify authentication
- Mobile responsive
- Use Tailwind CSS

Return production-ready code.

This gives the AI everything it needs.

Give the AI Context

Instead of saying:

Add login.

Try:

This project is a React + Tailwind application.

The authentication already exists.

Add a Forgot Password page that matches the existing design system.

Do not modify any authentication logic.

Only create the new page and routing.

Tell the AI What NOT to Change

This is one of the biggest mistakes users make. Instead of:

Improve this project.

Use:

Improve only the dashboard UI.

Do NOT modify:

- backend
- authentication
- routing
- database
- API endpoints

This keeps the AI focused.

Make Small Changes

Large prompts usually create larger code changes. Instead of:

Rewrite my application.

Break it into steps:

Improve navbar
Improve dashboard
Improve settings page
Improve mobile layout
Improve performance

Smaller prompts are easier to review and often require fewer credits.

Why Better Prompts Save Credits

Every AI task consumes credits based on the computing resources required to complete it. A detailed prompt helps the AI understand your request more accurately, which can reduce unnecessary iterations and make development more efficient. Poor prompts often require multiple attempts before reaching the desired result.

Example

❌ Poor Prompt

Improve my website.

Possible outcome:

AI guesses your intent
Multiple sections may change unexpectedly
Additional prompts are needed to refine the result
More credits may be consumed across repeated iterations

✅ Better Prompt

Improve only the homepage hero section.

Requirements:

- Keep existing branding
- Preserve navigation
- Improve typography
- Add a stronger call-to-action
- Do not modify any other pages

Possible outcome:

AI focuses only on the requested section
Fewer revisions
More predictable results
Reduced overall credit usage

Why Credit Usage Varies

Credit usage depends on the amount of work the AI performs, not simply the length of your prompt. Factors include:

Number of files modified
Complexity of the requested changes
AI model used
Images or videos generated
Code generation
External tools or integrations
Research and reasoning required

A short prompt can still consume more credits if it results in significant code generation or complex processing.

Best Practices

Be Specific

Describe exactly what should be changed instead of asking for broad improvements.

Protect Existing Work

Tell the AI which parts of your project should remain unchanged.

Break Large Tasks

Split large requests into smaller, focused prompts for better control.

Refine Instead of Restarting

Update your prompt based on the previous result instead of starting over whenever possible.

Use Existing Components

Instead of:

Create a pricing page.

Try:

Create a pricing page using the same design system already used in the dashboard.

Reuse existing buttons, typography, spacing, and color palette.

Ask the AI to Think Before Coding

A useful pattern is:

Before writing code:

Analyze the project.

Explain your implementation plan.

Identify affected files.

Then implement the solution.

This reduces unnecessary rewrites.

For Large Projects

If your application has many files: Tell the AI:

Read the existing project structure first.

Understand all related files.

Only modify files directly related to this feature.

Avoid changing unrelated code.

Working With Existing Code

Instead of:

Fix my app.

Use:

Investigate why the application fails.

Identify the root cause.

Explain it first.

Then make only the minimum required changes.

Image Generation Tips

Good prompt:

Modern fintech dashboard

Blue gradients

Minimal UI

Glassmorphism

Dark mode

4K

16:9

Professional SaaS style

Bad prompt:

Make image.

Video Generation Tips

Always specify:

duration
language
subtitles
aspect ratio
animation style
camera movement
background music
audience

Example:

Generate a 20-second promotional video.

Aspect ratio: 16:9

Language: English

Include subtitles.

Use smooth transitions.

Professional business style.

Keep text readable.

Background music should be upbeat.

Website Generation Tips

Example:

Create a landing page for an AI resume builder.

Tech stack:

- React
- Tailwind
- TypeScript

Pages:

- Home
- Pricing
- Blog
- Contact

Requirements:

- Dark mode
- SEO friendly
- Mobile responsive
- Accessibility compliant

Do not use placeholder text.

Mobile App Generation Tips

Example:

Create an Expo React Native application.

Requirements:

- Bottom navigation
- Authentication
- Offline support
- Light and Dark mode

Keep the folder structure clean.

Use TypeScript.

Do not use deprecated libraries.

Document Generation Tips

Instead of:

Create documentation.

Use:

Create Mintlify documentation.

Include:

- Overview
- Installation
- Examples
- API Reference
- FAQ
- Best Practices

Use MDX.

Include callouts and code blocks.

Before Pressing Generate

Did I explain the goal?
Did I mention my framework?
Did I tell the AI what NOT to change?
Did I include examples?
Did I explain the expected output?
Is my request specific?

Common Mistakes

Using very short prompts

Short prompts force the AI to guess your intent.Always include context and requirements.

Changing too many things at once

Break large tasks into smaller prompts.

Not specifying constraints

Tell the AI what should remain unchanged.

Regenerating repeatedly

Instead of clicking Generate again, improve your prompt.

Asking AI to rewrite an entire application

Large rewrites are harder to review and can introduce unnecessary changes.Modify one feature at a time.

AI Limitations

Even with excellent prompts, AI may occasionally:

Misinterpret instructions
Produce unexpected results
Require multiple iterations
Generate inefficient code
Miss edge cases

Always review generated output before publishing.

Best Practices

Give Context

Explain your project before asking for changes.

One Task at a Time

Small prompts usually produce better results.

Protect Existing Code

Tell the AI what should not change.

Use References

Screenshots, designs, and examples improve accuracy.

Review Before Publishing

Always test AI-generated changes.

Iterate

Improve prompts instead of regenerating repeatedly.

How Credits Work
AI Limitations
Website Hosting
Common Mistakes

​Prompt Engineering Guide

​How AI Understands Your Prompt

Context

Goal

Requirements

Constraints

Examples

Output

​The Anatomy of a Great Prompt

​Example: Poor Prompt

​Example: Great Prompt

​Give the AI Context

​Tell the AI What NOT to Change

​Make Small Changes

​Why Better Prompts Save Credits

​Example

​❌ Poor Prompt

​✅ Better Prompt

​Why Credit Usage Varies

​Best Practices

Be Specific

Protect Existing Work

Break Large Tasks

Refine Instead of Restarting

​Use Existing Components

​Ask the AI to Think Before Coding

​For Large Projects

​Working With Existing Code

​Image Generation Tips

​Video Generation Tips

​Website Generation Tips

​Mobile App Generation Tips

​Document Generation Tips

​Before Pressing Generate

​Common Mistakes

​AI Limitations

​Best Practices

Give Context

One Task at a Time

Protect Existing Code

Use References

Review Before Publishing

Iterate

​Related Guides

Prompt Engineering Guide

How AI Understands Your Prompt

The Anatomy of a Great Prompt

Example: Poor Prompt

Example: Great Prompt

Give the AI Context

Tell the AI What NOT to Change

Make Small Changes

Why Better Prompts Save Credits

Example

❌ Poor Prompt

✅ Better Prompt

Why Credit Usage Varies

Best Practices

Use Existing Components

Ask the AI to Think Before Coding

For Large Projects

Working With Existing Code

Image Generation Tips

Video Generation Tips

Website Generation Tips

Mobile App Generation Tips

Document Generation Tips

Before Pressing Generate

Common Mistakes

AI Limitations

Best Practices

Related Guides