Module 1: Exercise Workbook

Note: Scenario E would become agent-style if you said “save the outline options as a file.” A handy rule of thumb: “tell me / suggest” is chatbot-style, while “create / execute” is agent-style.

Task name: Creating a monthly sales report

[Parts humans should handle]
1. Deciding the report's purpose and approach
2. Interpreting the numbers and making business judgments
3. Final review and approval

[Parts that can be delegated to AI]
1. Importing and aggregating CSV data
2. Creating charts
3. Drafting the report document based on a template

Task name: Creating competitive analysis materials

[Parts humans should handle]
1. Defining the analysis perspective and comparison criteria
2. Drawing strategic insights
3. Leading the internal presentation and discussion

[Parts that can be delegated to AI]
1. Gathering competitor information via web search
2. Organizing information and creating comparison tables
3. Drafting the document

Key point: “Judgment,” “decision-making,” and “interpersonal communication” belong to humans; “gathering,” “organizing,” and “drafting” belong to AI. There’s no single right answer, so think about what fits your own work.

Autonomy: You just tell it what deliverable you want, and the AI figures out the steps and works through them on its own. Even when problems come up, it finds ways to handle them.

Tool use: It reads and writes files, searches the web, crunches numbers, and more — using a variety of tools to get work done. It’s like how you switch between Excel, a browser, and email throughout the day.

Multi-step execution: For requests like “gather the data, analyze it, and compile it into a report,” it can handle the entire multi-step workflow from start to finish in a single request.

Task 1: Creating weekly activity reports
  Current approach: Manually pull numbers from a spreadsheet, copy/paste into a document (about 2 hours)
  With Claude Code: Delegate everything from data aggregation to report generation in one request (about 10 minutes)

Task 2: Customer data cleansing
  Current approach: Manually filter and fix data in a spreadsheet (half a day)
  With Claude Code: Request "remove duplicates," "fix inconsistencies," and "standardize formats" all at once

Task 3: Formatting meeting minutes
  Current approach: Transcribe notes into a document and format them (30 min - 1 hour)
  With Claude Code: Hand over the raw notes and ask "organize these into meeting minutes format"

Key point: “Repetitive tasks,” “converting to a standard format,” and “processing large amounts of data” are areas where Claude Code excels.

Q: node -v says command not found A: Node.js is not installed. Refer to “3.3 Installing Node.js” in the textbook.

Q: npm install gives an error (Mac) A: Try sudo npm install -g @anthropic-ai/claude-code. When prompted for a password, enter your PC login password (nothing will appear on screen as you type — this is normal).

Q: claude command says command not found A: Close your terminal, reopen it, and try running claude again.

Q: The authentication page (browser) doesn’t open A: Try manually copying and pasting the URL shown in the terminal into your browser.

Results vary by environment, but should look roughly like this:

# 1. pwd result example (Mac)
/Users/jsmith

# 2. ls result example
Desktop    Documents    Downloads    Music    Pictures

# 3. mkdir test-folder
(No output means success)

# 4. ls result example
Desktop    Documents    Downloads    Music    Pictures    test-folder

# 5. cd test-folder
(No output means success)

# 6. pwd result example
/Users/jsmith/test-folder

What to check: In step 4, test-folder should appear in the list. In step 6, the path should end with test-folder.

Cleanup: To remove the folder you created in this exercise:

cd ..
rmdir test-folder

(cd .. means “go up one folder,” and rmdir means “remove directory”)

Example of a file Claude Code might create:

# Self-Introduction Card

## Basic Information

| Item | Details |
|------|------|
| Name | Jane Smith |
| Department | Marketing |

## Strengths

- Data analysis
- Project planning

## Hobbies

- Reading
- Running

What to check:

• The file was created
• A permission prompt appeared and the file was created after you approved
• It uses Markdown format (# for headings, - for bullet points)

Note: Claude Code’s output may vary each time. Even if the layout differs from the example above, it’s fine as long as the content is there.

What makes instructions better:

• Who/what for: Specify the purpose and audience
• What: Describe the specific content
• In what format: Specify the file format and structure
• Where: Specify the output destination and file name

Results will vary depending on your environment and version, but you should observe the following:

• /help: Displays a list of commands with brief descriptions
• /cost: Shows an amount like $0.XX (a small amount is normal)
• /clear: Shows a message like “Conversation cleared” and the previous conversation disappears
• /exit: Claude Code exits and you return to the terminal prompt ($ or %)

Note: Running /clear resets the conversation context. Avoid using it in the middle of a task.

Example of a generated CLAUDE.md:

# Project Configuration

## Team Member Info
- Department: Strategic Planning
- Main tasks: Business planning, budget management, executive meeting materials

## Work Rules
- Create all files in English
- File names follow the format "YYYY-MM-DD_description"
- Reports are created in Markdown format (.md)

## Output Format
- Use comma separators for numbers (thousands)
- Dates in YYYY-MM-DD format
- Include currency symbols with monetary amounts

What to check:

• CLAUDE.md was created in the current directory
• Your work details are reflected
• The necessary rules are documented

Expected differences:

Key takeaway: With CLAUDE.md in place, your instructions can be more concise and the output quality is more consistent. Hopefully you’ve experienced that “CLAUDE.md takes a bit of upfront effort but saves significant time in the long run.”

Example of the updated CLAUDE.md:

# Project Configuration

## Team Member Info
- Department: Strategic Planning
- Main tasks: Business planning, budget management, executive meeting materials

## Work Rules
- Create all files in English
- File names follow the format "YYYY-MM-DD_description"
- Reports are created in Markdown format (.md)

## Output Format
- Use comma separators for numbers (thousands)
- Dates in YYYY-MM-DD format
- Include currency symbols with monetary amounts

## Folder Structure
- `data/` : Source data (CSV and spreadsheet files)
- `reports/` : Completed report output
- `drafts/` : Drafts and work-in-progress files

## Common Requests
1. Aggregate CSVs from data/ and create a monthly report
2. Create summary materials for executive meetings
3. Create budget vs. actual variance analysis reports

## Important Notes
- Do not create files containing personal information (names, addresses, phone numbers, etc.)
- Do not include confidential terms in file names
- All monetary figures must be verified by a human
- Documents intended for external distribution must be created as drafts for human review

Key point: CLAUDE.md is something you “grow over time.” As you use it, whenever you think “I need this rule too” or “this explanation is missing,” keep adding to it.

Post-completion checklist:

[  ] CLAUDE.md has been created
[  ] data/ folder exists
[  ] reports/ folder exists
[  ] templates/ folder exists
[  ] templates/ contains a report template
[  ] data/ contains a sample CSV
[  ] reports/ contains an analysis report
[  ] The contents of each file are as expected

How to verify via terminal (after exiting Claude Code):

# Check the folder structure
ls -la ~/my-first-project/
ls -la ~/my-first-project/data/
ls -la ~/my-first-project/reports/
ls -la ~/my-first-project/templates/

Reflection points:

• Through the 5 requests, Claude Code executed a complete workflow: file creation -> data generation -> analysis -> report creation
• Permission checks were appropriately triggered
• CLAUDE.md rules were reflected (e.g., file naming format)
• The cost shown by /cost was within a reasonable range