Skip to main content
Version history

Part 2: Agent Workflow Primitives

You understand the paradigm. You know the Seven Principles. You've seen what an AI Employee can do (Chapter 7).

Now it's time to build the skills to create your own.

This part is a progressive skills lab. Each chapter tackles a specific problem domain — organizing files, synthesizing research, analyzing data, generating documents, and managing versions. By Chapter 13, you combine everything to build your own AI Employee.

The Progressive Arc​

Each chapter solves a real-world problem while explicitly applying the Seven Principles from Chapter 6.

Two Interfaces, One Agent​

Throughout this part, you'll use your General Agent through two interfaces:

  • Code (Claude Code): For precise operations — file processing, data analysis, version control, automation scripts
  • Cowork (Claude Desktop): For iterative work — research synthesis, document generation, planning strategies

Each chapter specifies which interface suits the problem. By Chapter 12, you'll fluidly switch between both.

A Note on Mindset​

The most important thing you'll develop in this part isn't tool proficiency — it's the ability to decompose real problems into agent-solvable tasks.

Each chapter starts with a problem you recognize from your own work. The solution always follows the same pattern:

  1. Describe what you need (not how to do it)
  2. Verify the agent understood correctly
  3. Execute in small, reversible steps
  4. Persist the output for future reuse

This is the General Agent workflow. By Chapter 13, it will be second nature.