AGENTIC CODING WITH CLAUDE CODE

Move past prompt-stitching by mastering the engineering disciplines that make Claude Code reliable in production codebases.

Key Features
- Master the engineering disciplines that make Claude Code reliable in production codebases - not just demos

- Apply the four-surface model and the PPAO loop to diagnose any agent behavior in seconds

- Build a governance posture that survives security review - OWASP LLM Top 10, NIST AI RMF, ISO/IEC 42001, EU AI Act, SOC 2

Book Description

The agent reads your codebase. It plans the change. It writes the tests. It runs them and reports "all passed." Everything looks fine - until you notice the tests passed because the agent quietly disabled the one that would have failed.

If you've used Claude Code for any serious length of time, you've had a version of this moment.

This book is what you reach for next.

Thirteen chapters of working disciplines - specification, verification, attribution, governance, team adoption - written for senior engineers and engineering leaders who need agentic coding to hold up in production. Not a feature tour. Not a list of prompts. Not an argument that AI is great. A working manual for the people responsible for making this work.

You'll learn how the agent actually thinks, how to specify tasks it can execute reliably, how to verify its work at production-grade rigor, and how to debug failures by reading traces diagnostically. You'll learn how to extend Claude Code with MCP, govern its participation in version control and CI/CD, and place the human-in-the-loop boundary where it actually catches mistakes. And you'll learn how a team sustains agentic-coding adoption past the early-adopter wave - through role structure, measurement, and the lived contract of CLAUDE.md.

The tool will change. These disciplines won't.


What You Will Learn

• - Apply the PPAO loop and the four surfaces (Tools, Context, Permissions, Memory) to predict and diagnose any agent behavior
• - Write specifications the agent can execute reliably using the five-element template
• - Verify agent-generated code at production-grade rigor with the four-check framework
• - Read agent traces diagnostically using the six-class failure taxonomy
• - Extend Claude Code with MCP servers - adoption, design, and the trust posture each warrants
• - Govern agent participation in Git, CI/CD, and production deployment workflows
• - Map team controls to OWASP LLM Top 10, NIST AI RMF, ISO/IEC 42001, EU AI Act, GDPR, HIPAA, PCI-DSS, FedRAMP, and SOC 2
• - Sustain team adoption with the four-role structure and Goodhart-resistant measurement