Running Claude Code and Crafting Prompts

Working with AI agents means giving clear instructions. Just like a new team member, an AI needs to understand what you want to achieve. By the end of this lesson you'll be able to launch the Claude Code environment and construct effective prompts using the 'Take → Process → Store' framework.

Core idea

To interact with an AI agent like Claude Code, you'll use a prompt. A prompt is a text instruction that tells the AI what task to perform. The clearer and more precise your prompt, the better the AI's output will be. Think of it as giving directions to a smart assistant: vague instructions lead to vague results.

Claude Code operates within a terminal environment. This is a text-based interface where you type commands and receive responses. To start working, you first need to navigate to your project folder and then launch the Claude Code application.

We'll use a simple, yet powerful, framework for crafting prompts: Take → Process → Store. This helps you structure your instructions logically:

• Take: What information or data are you giving the agent? This could be a file, a folder, a link, or just raw text.
• Process: What do you want the agent to do with that information? Examples include finding patterns, comparing items, summarizing, or restructuring data.
• Store: What should the agent produce as an output, and where should it put it? This is crucial for getting the results in a usable format, such as a specific file type or location. If you don't specify where to store the output, the agent will usually display it directly in the terminal.

Walkthrough

Let's launch Claude Code and then craft a prompt to analyze a meeting transcript.

Task: Launch Claude Code and analyze a meeting transcript.

1. Open your terminal. This is the command-line interface on your computer.
2. Navigate to your project folder. Your project files are typically located in a directory named datapeople.

Press Enter after typing the command. The `cd` command stands for "change directory," and `~/datapeople` refers to your `datapeople` folder within your home directory.

3. Launch Claude Code.

Press Enter. You should now see the Claude Code prompt, indicating it's ready for your instructions.

4. Enter the following prompt into the Claude Code environment:

This prompt, translated to English, instructs Claude Code to:
*   **Take:** The file `calls/2026-03-03_weekly-sync.md`, which contains a meeting transcript.
*   **Process:** Read the entire transcript, identify key decisions, open questions, and problems discussed. For each item, specify what it is, who mentioned it, and its context.
*   **Store:** (Not explicitly specified, so Claude Code will display the output in the terminal.)

Common mistakes

• Forgetting to launch Claude Code: You must be in the Claude Code environment to interact with the AI. If you just type claude and it doesn't work, ensure you are in the correct project directory first.
• Vague prompts: If your prompt isn't specific about what to "Process" or "Store," the AI might give you a generic or unhelpful output. Always aim for clarity.
• Incorrect file paths: Double-check that any file or folder paths you provide in your prompt are accurate. A small typo can prevent the AI from finding the data.

Key takeaways

• Prompts are text instructions for AI agents.
• Launch Claude Code by navigating to your project folder and typing claude in the terminal.
• Use the Take → Process → Store framework to structure your prompts.
• Specify what data the AI should Take, what it should Process, and where it should Store the output.
• Clear and specific prompts lead to better AI results.