Better Operators

The Art of Prompting

Good prompts are the difference between getting useful code and getting frustrated. These strategies work across all AI coding tools.

Strategy 1: Be Specific About Context

Don't assume the AI knows your setup. State your tech stack, constraints, and goals explicitly.

Bad vs Good

❌ Bad: "Create a login form"

✅ Good: "Create a login form using React and Tailwind CSS.
Requirements:
- Email and password fields
- Client-side validation
- Show loading state during submit
- Handle errors with toast notifications
- Use our existing Button component from @/components/ui"

Strategy 2: The Q&A Strategy

Get the AI to ask clarifying questions before providing solutions. This helps the model understand your requirements better.

Q&A prompt template

I need to implement Stripe payments. Before providing a solution, please ask me relevant questions about my specific requirements and constraints so you can give me the most appropriate implementation advice.

Best for:

API integrations
Authentication systems
Database design
Complex app features (notifications, etc.)

Strategy 3: Pros and Cons Analysis

Prevent one-sided recommendations by asking for balanced analysis.

Pros/Cons prompt template

I need to implement transactional email into my app. Analyze the pros and cons of Resend, Loops, and Mailersend. Consider factors like scalability, UX, ease of development, and maintenance requirements.

Strategy 4: Role-Based Prompting

Get advice that reflects a specific role's priorities and expertise.

Role-based prompt template

You're a security expert and ethical hacker. Review my app to identify any security vulnerabilities, potential edge cases, or implementation flaws. Be particularly attentive to common OWASP security risks.

Strategy 5: Provide Examples

Show the AI what you want by providing examples of similar code in your codebase.

Example-driven prompting

Create a ProductCard component following the same pattern as this existing UserCard:

[paste UserCard code]

The ProductCard should display: image, title, price, and an 'Add to Cart' button.

Strategy 6: Iterative Refinement

Don't try to get everything perfect in one prompt. Build incrementally.

The iteration loop:

1. Start with core functionality
2. Review and test what you got
3. Ask for specific modifications
4. Add edge cases and polish
5. Refactor if needed

Strategy 7: Constrain the Output

Tell the AI what NOT to do. This prevents over-engineering.

Using constraints

Create a simple counter component.

Constraints:
- No external state management (no Redux/Zustand)
- No TypeScript generics
- Keep it under 30 lines
- Don't add features I didn't ask for

The best prompt is one you can reuse. Save your effective prompts as templates.