Beginner

Prompt Engineering Fundamentals

Master the building blocks of effective prompts: anatomy, clarity, role assignment, output constraints, and the CRISP framework.

The Anatomy of a Prompt

Every effective prompt consists of four key components. You do not always need all four, but understanding each one helps you construct better prompts:

Prompt Anatomy

1. INSTRUCTION - What you want the AI to do
   "Summarize the following article..."

2. CONTEXT - Background information the AI needs
   "You are reviewing a technical paper for a
    non-technical audience..."

3. INPUT - The data or content to process
   "Article: [paste article text here]"

4. OUTPUT FORMAT - How you want the response structured
   "Provide a 3-paragraph summary with bullet points
    for key takeaways."

Clarity and Specificity

The most common mistake in prompt writing is being too vague. AI models respond to exactly what you ask, not what you mean. Here is how to be specific:

Vague Prompt	Specific Prompt	Why It's Better
"Write about dogs"	"Write a 500-word blog post about the top 5 dog breeds for apartment living, including size, temperament, and exercise needs"	Defines topic, length, structure, and content
"Fix this code"	"This Python function should return a sorted list but returns None. Find the bug, explain why it happens, and provide the corrected code"	States expected behavior, asks for explanation
"Help with email"	"Write a professional email to a client explaining a 2-week project delay. Tone should be apologetic but confident. Include a revised timeline"	Specifies audience, tone, and required content

Role Assignment

Assigning a role to the AI model is one of the most powerful prompt engineering techniques. When you tell the AI to act as a specific expert, it draws on knowledge patterns associated with that role.

Role Assignment Examples

// Technical reviewer
"You are a senior software engineer with 15 years of
experience in Python. Review this code for performance
issues, security vulnerabilities, and best practices."

// Creative writing coach
"You are a creative writing professor at a top MFA
program. Provide constructive feedback on this short
story, focusing on character development, pacing,
and dialogue authenticity."

// Data analyst
"You are a data analyst at a Fortune 500 retail
company. Analyze this sales data and provide insights
about seasonal trends, top-performing products, and
recommendations for inventory planning."

✅

Tip: Be specific about the role's experience level and domain. "A senior Python developer" produces different results than "a junior developer learning Python." The role shapes the vocabulary, depth, and assumptions in the response.

Tone and Style Control

You can control the tone and writing style of AI responses by being explicit about your expectations:

Formal: "Use professional, academic language suitable for a journal publication."
Casual: "Write in a friendly, conversational tone as if explaining to a friend."
Technical: "Use precise technical terminology. Assume the reader has a CS degree."
Simple: "Explain this as if teaching a 10-year-old. Avoid jargon entirely."
Persuasive: "Write in a compelling, persuasive style suitable for a sales page."

Output Constraints

Constraining the output format ensures you get responses that are immediately useful. Here are the key constraints you can set:

Length

"Keep your response under 200 words" or "Write exactly 5 bullet points" or "Provide a one-paragraph summary"

Format

"Respond in JSON format" or "Use a markdown table" or "Format as a numbered list" or "Use XML tags"

Language

"Respond in Spanish" or "Use British English spelling" or "Write at an 8th-grade reading level"

Structure

"Include an introduction, three body sections, and a conclusion" or "Start each section with a question"

Good vs Bad Prompt Comparison

Let us compare prompts side by side to see the principles in action:

Example 1: Code Generation

// BAD: Vague, no context, no constraints
"Make a login form"

// GOOD: Specific technology, features, and requirements
"Create a React login form component with:
- Email and password fields with validation
- Show/hide password toggle
- 'Remember me' checkbox
- Error message display for invalid credentials
- Loading state on submit button
- Use TypeScript and Tailwind CSS
- Include aria-labels for accessibility"

Example 2: Content Writing

// BAD: No audience, no purpose, no structure
"Write about climate change"

// GOOD: Audience, purpose, structure, tone defined
"Write a 1000-word article for a high school science
magazine about how climate change affects ocean
ecosystems. Include:
- An engaging hook in the first paragraph
- 3 specific examples with data
- A section on what students can do
- Tone: educational but not preachy
- Reading level: 10th grade"

The CRISP Framework

CRISP is a mnemonic framework for constructing well-structured prompts. Use it as a checklist when writing important prompts:

Letter	Component	Description	Example
C	Context	Background information and situation	"I'm a startup founder preparing for a Series A pitch"
R	Role	Who the AI should act as	"Act as an experienced venture capital advisor"
I	Instructions	The specific task to perform	"Review my pitch deck outline and provide feedback"
S	Specifics	Details, constraints, requirements	"Focus on narrative flow, market size claims, and financial projections"
P	Parameters	Output format, length, style	"Provide feedback as a numbered list with severity ratings (critical/important/nice-to-have)"

CRISP Framework in Action

Context: I'm a startup founder building a B2B SaaS
product for HR teams. We have 50 paying customers
and $200K ARR. Preparing for Series A fundraising.

Role: Act as an experienced VC partner who has
evaluated hundreds of Series A pitches in the
enterprise SaaS space.

Instructions: Review the pitch deck outline below
and provide actionable feedback to strengthen it.

Specifics: Focus on: (1) whether the narrative arc
is compelling, (2) if our market sizing is credible,
(3) gaps that VCs will question, (4) slides that
should be added or removed.

Parameters: Format your feedback as a table with
columns: Slide, Issue, Severity, Recommendation.
Then add a summary paragraph with your overall
assessment.

💡

Remember: You do not need to use all five CRISP components every time. For simple tasks, just Instruction and Specifics may suffice. Use the full framework for complex, high-stakes prompts where precision matters.

← Previous Introduction Next → Techniques

Ready to Go Deeper?

Live instructor-led courses from our partners. Affiliate disclosure.

🧠

AI & ML Courses - 30% Off

Live instructor-led AI, machine learning, data science, and cloud courses for working professionals. Use code Limited30 at checkout.

Edureka

📊

DataCamp - AI & Data Science

Hands-on Python, machine learning, and AI courses with interactive exercises and real projects.

DataCamp

🎓

edX - Top AI Courses

University-level AI courses from MIT, Harvard, Stanford. Earn certificates that employers recognize.

edX