Hari Krishnan

Better agent-produced software starts long before the first line of code, and confidence requires more than a passing test suite. This ebook brings together four expert chapters on expressing intent, writing durable specifications, testing according to risk, and evaluating code you didn’t write. It’s an excellent starting point for anyone ready to replace guesswork with a practical foundation for directing agents, reviewing their work, and building dependable software faster.

For AI coding agents, accurately specifying your intent is the whole ballgame! Spec-Driven Development: Engineering with intent teaches you how to close the gap between what you mean and what your agents actually build. In this practical, tool-agnostic playbook, author Hari Krishnan shows you how to make your intent clear through structured specifications and then make those specs the primary control surface for your projects, rather than the code itself.

This book is hands-on from the very beginning. You’ll follow a running URL shortener application that grows in ambition from chapter to chapter. As you explore the intent-first spec-driven development model, you’ll learn to adjust your process for differences in alignment, human and agentic collaboration, and evolving outcomes. As the project scales into microservices, multiple repositories, and cross-team coordination, you’ll see how intent-led development pays out with infrastructure, deployment, security, acceptance testing, and architectural governance.

Everything you’ll read and do is grounded in established software engineering fundamentals—extreme programming, legacy code refactoring, technical debt management—so you can connect new AI workflows to what you already know. You’ll see how to drive defect fixes through specifications rather than direct code edits, how to apply SDD to brownfield codebases with no existing specs, and how to construct an intent-harness that continually improves its own workflows from implementation feedback.

Hands-on exercises, including team-based collaboration through a shared GitHub repository, let you practice the workflows as they’d play out on a real engineering team. A companion code repository with full commit history lets you follow every step of the implementation rather than just seeing the finished result.