The Anatomy of a Great Prompt.
Five building blocks. Infinite combinations. One framework that works every time.
After this lesson you'll know
- The 5 components of every effective prompt
- Which components are mandatory vs optional
- How to assemble them in the right order
- A reusable template you can apply to any task
The Framework
The RCFCE Framework.
Every great prompt is built from five components. You don't always need all five, but knowing them gives you a toolkit for any situation.
Priorities
You don't always need all five.
Here's the truth: for most everyday prompts, you need Context + Format at minimum. Those two alone put you ahead of 90% of AI users.
Always Include
- ✓ Context — what you need and why
- ✓ Format — what the output should look like
Add When Needed
- + Role — for specialized expertise
- + Constraints — for precision
- + Examples — for style matching
Practice
Let's build a prompt piece by piece.
Say you need help writing a thank-you email to a client. Watch how each layer improves the result:
Just the ask
"Write a thank you email."
+ Context
"Write a thank you email to a client who just signed a $50K annual contract with our design agency. They chose us over two larger competitors."
+ Context + Format + Constraints
"Write a thank you email to a client who just signed a $50K annual contract with our design agency. They chose us over two larger competitors. Keep it under 150 words. Tone: warm and professional, not salesy. End with a concrete next step (scheduling a kickoff call). No exclamation marks."
+ Role + Context + Format + Constraints
"You're the founder of a 12-person design agency. Write a thank you email to a client who just signed a $50K annual contract. They chose us over two larger competitors. Keep it under 150 words. Tone: warm, confident, and founder-to-founder — not corporate. End with a concrete next step (scheduling a kickoff call). No exclamation marks."
Dissection
Breaking down prompts component by component.
The best way to internalize the RCFCE framework is to see it deconstructed in real prompts. Let's dissect three prompts and label every component.
Dissection 1 — A Blog Post Prompt
[R] You are a tech journalist who writes for Wired.
[C] Write a 600-word blog post about why most AI tools
fail in enterprise companies. My readers are CTOs and
VP Engineering at companies with 500+ employees. They
are skeptical of AI hype and want practical insight.
[F] Structure: hook (1 paragraph) → 3 reasons with subheads
→ closing with one actionable takeaway. No bullet points.
[C] Tone: sharp, slightly provocative, data-informed. Under
600 words. Do not use "game-changing," "revolutionary,"
or "in today's rapidly evolving landscape."
[E] Style reference: write like Ben Thompson (Stratechery)
— analytical, opinionated, no fluff.
ALL 5 components present. This prompt leaves almost nothing
to chance. The AI knows who to be, who to write for, what
format to use, what boundaries to respect, and what good
looks like. The output will be publication-ready.Dissection 2 — A Data Analysis Prompt
[R] You are a data analyst presenting to non-technical
executives.
[C] Our Q3 website traffic dropped 23% compared to Q2.
We launched a redesign in July. We also cut our ad
budget by 40% the same month. Help me figure out
which factor drove the decline.
[F] Format: Start with your best hypothesis in 1 sentence.
Then give me a table with columns: Factor, Evidence For,
Evidence Against, Confidence Level (High/Med/Low).
[C] Do not hedge with "it could be many things." Give me
your actual assessment. Under 300 words total.
[E] — Not included, and that is fine for this task.
The context and format are strong enough to produce
a focused, actionable analysis.Dissection 3 — A Simple Prompt That Works
[C] I need to tell my team of 12 that our Friday deadline
is moving to next Wednesday.
[F] Write a 3-sentence Slack message.
[C] Encouraging, not apologetic.
No role. No examples. Just Context + Format + one Constraint.
And it will work perfectly — because the task is simple.
THIS IS THE KEY INSIGHT: Match your prompt complexity to
your task complexity. Not every prompt needs all 5 components.These dissections reveal the real skill: knowing which components to include based on the task. Complex, high-stakes work needs all five. Quick everyday tasks need two or three. The framework gives you options — not obligations.
Impact Map
What each component actually changes in the output.
Understanding the RCFCE components is one thing. Understanding what each one does to the output helps you decide which ones to invest time in.
Role changes: vocabulary, depth of expertise, perspective, confidence level. A "junior copywriter" hedges. A "senior conversion specialist" commits. The same knowledge, filtered through a different lens.
Context changes: relevance, specificity, usefulness. Without context, AI gives you information that is correct but generic. With context, it gives you information that is correct AND applicable to your specific situation.
Format changes: structure, scanability, usability. The difference between a wall of text you have to reformat and something you can paste directly into your deliverable. Format is about respecting YOUR time.
Constraints change: precision, authenticity, focus. They eliminate the generic, the verbose, and the AI-sounding. Constraints are what turn "pretty good" output into "actually usable" output.
Examples change: style, voice, quality bar. They show AI exactly what "good" looks like for YOU. Two sentences of example can do more for style matching than a paragraph of description.
Your Template
Copy this. Use it everywhere.
[ROLE] You are a [expertise/persona]. [CONTEXT] I need help with [situation]. Background: [relevant details]. Audience: [who will see this]. [FORMAT] Give me [specific output format]. [CONSTRAINTS] - [Length/word count] - [Tone] - [Things to avoid] - [Specific requirements] [EXAMPLE] (optional) Here's an example of what I'm looking for: [paste example]
You don't need to use labels like [ROLE] in your actual prompts — those are just training wheels. Once you internalize the framework, you'll naturally include these elements in flowing, natural language.