Teacher’s Edition
Module 02
Prompt Anatomy
The five-part prompt structure: Role, Context, Task, Constraints, Format. Sixty seconds longer to write. Six fewer rounds of revision. The single biggest skill jump in the program.
Charter Oak Strategic Partners · Claude Mastery Program · Version 1.0 · Confidential · Not for distribution to participants
The five-part prompt structure (Role, Context, Task, Constraints, Format) is not arbitrary. Each component addresses a different failure mode. Without Role, Claude defaults to a generic assistant voice. Without Context, it lacks the facts it needs to be specific. Without Task, it guesses what you want. Without Constraints, it over-produces or includes irrelevant material. Without Format, it picks a shape that may not match your needs.
Most people naturally include the Task. Some remember Context. Almost nobody provides Role, Constraints, or Format until they are taught to. The discipline of including all five is the single largest skill improvement in this program. It sounds simple. The gap between knowing it and doing it consistently is where the training earns its value.
Opening — 5 minutes
“Every good prompt has five parts. Role: who is Claude in this conversation? Context: what does Claude need to know? Task: what exactly should Claude do? Constraints: what should Claude avoid, and what are the boundaries? Format: what should the output look like?”
“That is the whole framework. Five lines. You can write them in sixty seconds. And they will transform every interaction you have with Claude from this point forward.”
“Take the worksheet in front of you. We will fill one out together, then you will do it on your own.”
demo-data/module-02/prompt-template-worksheet.md— Fill-in-the-blank worksheet with self-check.Role. Tells Claude what expertise to bring. “You are a senior financial analyst” produces different language, assumptions, and depth than “You are a marketing intern.” The role also anchors tone. A legal compliance officer writes differently than a brand copywriter.
Context. The facts Claude needs. Company size, industry, situation, relevant history, data points. Claude cannot look things up. It works only with what you provide. More context produces more relevant output. The most common mistake is assuming Claude knows things about your organization that it cannot.
Task. The specific action. “Analyze” is vague. “Identify the three largest cost drivers in this dataset and rank them by percentage of total spend” is specific. Specificity here directly determines output quality.
Constraints. Boundaries. Maximum length, things to exclude, level of technicality, audience assumptions. “Keep it under 500 words” or “Do not recommend any solution that costs more than $50,000” or “Assume the reader has no technical background.” Constraints prevent Claude from over-producing or missing the mark.
Format. The shape of the output. Bullet points, narrative paragraphs, a table, an email, a slide outline. Claude will pick a format if you do not. It will not pick the one you wanted.
Live Demo: The Vague Email — 15 minutes
demo-data/module-02/vague-request-emails.md— Four vague emails with analysis and structured rewrites.“I want to show you something that happens in every organization. You get an email from a colleague. It says: ‘Hey, can you help me put something together about our new product for the leadership meeting next week?’”
“Let us see what happens when we paste that into Claude exactly as written.”
Paste Email 1 from the demo file. Wait for Claude’s output. It will be generic. A broad outline. No specifics. No voice. No useful detail.
“Now watch what happens when we rewrite it using the five components.”
Paste the structured five-part version from the demo file. The output will be specific, actionable, and formatted for the stated audience.
“How long did it take to write the structured version compared to the vague one?”
Let someone answer. Then:
“Maybe an extra minute. Was that extra minute worth six rounds of ‘no, that is not what I meant, try again’?”
The demo file contains four emails, each from a different department. Email 1 is a marketing request with no specifics about audience, format, or scope. Email 2 is an HR request that says “we need to do something about onboarding” with no clarity on what “something” means. Email 3 is a finance request for “a summary of our expenses” with no time range, no level of detail, and no audience specified. Email 4 is a sales request for “competitive intelligence on that company we keep losing deals to” with no company named and no definition of what intelligence would be useful.
Each email’s rewrite in the demo file identifies exactly what was missing and provides a complete five-component version. The rewrites are good teaching examples because they show how much specificity a human must add before Claude can do meaningful work.
Group Exercise: Transform the Email — 30 minutes
“Four groups. Each group gets one of the four emails. Your job: rewrite it as a five-part prompt using the worksheet. You have twelve minutes. Then test your prompt in Claude. See what you get.”
“After testing, each group presents. Read the original email. Read your prompt. Show Claude’s output. The room votes on which group produced the most usable result.”
“This is a competition. The group with the best output wins. What that means is: specificity wins. The more precise your prompt, the better your result.”
Walk the room during the exercise. The most common failure is a Task section that reads “analyze this” or “help with this.” Push them: “Analyze what, specifically? What does the output need to include? What decisions will someone make based on this?”
The second most common failure is omitting Constraints entirely. Ask each group: “Who reads this? How long should it be? What should it leave out?”
“What was the hardest part of the five components to write?”
Pattern: most groups struggle with Constraints and Role. Context is easier because the email gives them facts. Task is natural. Thinking about what to exclude and what perspective to take requires a different kind of thinking.
Individual Practice — 15 minutes
“Now do it for yourself. Think of a task you do at work. Something recurring. Something that takes you at least 30 minutes. Fill out the worksheet for that task. All five components. Then test it in Claude.”
“This prompt becomes the first entry in your Prompt Library, which we will build in Module 04. So pick something worth keeping.”
Some participants will say “I can’t think of anything.” Direct them to the Workflow Ideas by Role sheet from Module 04’s demo data. It has 25 examples across five departments. The goal is to get them writing, not to find the perfect task.
Debrief — 10 minutes
“What was your task, and which of the five components made the biggest difference in the output?”
Listen for the pattern. It will emerge: Context and Constraints are almost always the answer. People forget to tell Claude what it needs to know (Context) and what it should avoid (Constraints). Naming this pattern aloud is the teaching moment. It sticks because the participants discovered it themselves.
“The five components are your new default. Role, Context, Task, Constraints, Format. Every prompt. Every time. It takes sixty seconds longer. It saves you from the ‘no, that is not what I meant’ loop. And it is the foundation for everything we build in the rest of this program.”
| Segment | Activity | Time |
|---|---|---|
| Opening | Five components, distribute worksheet | 5 min |
| Demo | Vague email vs. structured prompt | 15 min |
| Group Exercise | Transform the email (4 groups) | 30 min |
| Individual | Build a prompt for your real work | 15 min |
| Debrief | Share and discuss | 10 min |