CIS 4399 · AI Prompting Playbook

Prompt Like a
Pro

A practical field guide to getting the best results from Generative AI — built on real experience across five AI platforms.

Claude ChatGPT AI Studio Gemini Copilot
Scroll to explore
Five Principles

The Framework That Works

These five principles — from Dr. Villacis Calderon's CIS 4399 — are the foundation of effective AI prompting. Apply them in order for best results.

01
🎯
Give Direction
  • Assign a role or persona to the AI
  • Use prewarming: ask what it knows first
  • Add context about your audience
  • Provide clear Do's and Don'ts
02
📐
Specify Format
  • Define output type: list, table, code, doc
  • Set length: word count, # of bullets
  • Specify tone and target audience
  • Add constraints: no jargon, include steps
03
💡
Provide Example
  • Zero examples = zero-shot (risky)
  • 2–3 examples = few-shot (recommended)
  • Examples reduce AI assumptions
  • Give direction first, then examples
04
🔍
Evaluate Quality
  • Never deploy without reviewing output
  • Test the right model for your task
  • Add quality guidelines to your prompt
  • Keep a human in the review loop
05
⚙️
Divide Labor
  • Don't overload one prompt with everything
  • Split tasks into focused sub-prompts
  • Use separate system prompts per role
  • Shared or isolated context — choose wisely
Know Your Tools

Match the Model to the Mission

Different AI models have different strengths. Choosing the right one is part of the Divide Labor and Evaluate Quality principles.

Claude
Best for: Writing & Code
Exceptional long-document analysis, coding, and nuanced writing. Strong at following complex instructions precisely.
ChatGPT
Best for: General Tasks
Versatile all-rounder. Strong at conversation, brainstorming, summarization, and iterative refinement.
AI Studio
Best for: Multimodal
Google's model playground. Great for experimenting with Gemini models and testing prompts at scale.
Gemini
Best for: Research
Deep Google integration. Ideal for research, real-time information retrieval, and Google Workspace tasks.
Copilot
Best for: Microsoft 365
Embedded in Word, Excel, and Teams. Best when working inside Microsoft's ecosystem for productivity tasks.
Before & After

Bad Prompt vs. Better Prompt

See how applying the five principles transforms a vague request into a prompt that gets results. All examples are based on building a website.

✗ Weak Prompt
"Make me a website."
Why it fails: No direction, no format, no context, no audience. The AI has to guess everything — and it will guess wrong.
→
Apply the
5 principles
✓ Strong Prompt
Direction + Format + Example "Act as a professional web developer. Create a single-file HTML portfolio website for a U.S. Army Signal Corps veteran and IT student. The site should include sections for: Hero, About, Experience, Skills, Certifications, and Contact. Use a dark military color scheme with olive green and tan. Return complete, working HTML/CSS/JS in one file. Make it mobile-responsive. Here is an example of a section heading style I like: bold, monospace font, with a colored underline accent."
Why it works: Assigns a role, specifies format (single HTML file), lists content requirements, defines visual direction, and includes an example of desired style. Nothing is left to guesswork.
✗ Weak Prompt
"Fix the bug on my website."
Why it fails: No description of the bug, no code provided, no expected behavior defined. The AI cannot help without information.
→
Apply the
5 principles
✓ Strong Prompt
Direction + Quality + Context "I have a bug in my HTML/CSS portfolio website. The custom cursor element is not appearing on screen, and sections 2 through 6 are invisible when the page loads. The cursor is defined as a div with id='cursor' using position:fixed, and the sections use a fade-in animation class that starts at opacity:0. Please diagnose the likely causes and provide a corrected version of the JavaScript scroll-reveal and cursor logic. Do not change any other part of the code."
Why it works: Describes the exact symptoms, provides technical context, specifies what not to change (a constraint), and asks for a clear diagnosis and fix. This is how professional debugging prompts are structured.
✗ Weak Prompt
"Write content for my website."
Why it fails: No tone, no audience, no structure, no word count. The output will be generic, off-brand, and unusable.
→
Apply the
5 principles
✓ Strong Prompt
Direction + Format + Tone + Example "Write the About section copy for my personal portfolio website. The brand voice is: technical, direct, and grounded — like a field manual, not a corporate brochure. The audience is IT recruiters and hiring managers. Keep it under 120 words. Write in first person. Here is an example of the tone I want: 'I've spent my career where signal meets steel — building systems that can't fail and leading teams that don't quit.'"
Why it works: Defines the brand voice, identifies the audience, sets a word limit, specifies perspective, and provides a tone example. The AI now has everything it needs.
Quick Reference

Prompting Cheat Sheet

Apply these practical tips immediately to level up any prompt you write.

// TIP 01
Start with a Role
"Act as a [role]" sets the AI's mindset before it generates anything. A web developer, a recruiter, a brand strategist — the role shapes everything.
// TIP 02
Define the Output
Always specify: "Return as a single HTML file" or "Give me 5 bullet points" or "Output a table." If you don't define it, AI will guess.
// TIP 03
Provide 2–3 Examples
Few-shot prompting dramatically improves accuracy. Show the AI what good output looks like and it will match that pattern.
// TIP 04
Always Review Output
Never copy-paste AI output without reading it. Evaluate it against your goal. AI makes confident mistakes — you are the quality check.
// TIP 05
Break Big Tasks Down
Don't ask AI to "build my entire website" in one prompt. Split it: brand guide first, then layout, then content, then debugging. Each prompt = one clear job.
// TIP 06
Name What Not to Do
Constraints are as powerful as instructions. "Do not change the navigation," "No jargon," "Do not use purple" — clear guardrails prevent unwanted changes.

Ready to Put It Into Practice?

Apply these principles on your next AI project. The difference between a frustrating prompt and a powerful one is context, clarity, and iteration.

Back to Portfolio