Context & Agent Engineering for Builders

Direct AI. Don't Just Chat With It.

You learned how to prompt. Now learn how to direct. Context & Agent Engineering for Builders teaches the two skills that separate occasional AI users from people who ship with AI: choosing what the model sees, and directing autonomous AI runs without losing control. Tool-agnostic, evergreen, written for developers and technical professionals who direct AI rather than chat with it.

Inside this book

• Context engineering: choosing what the model sees and what it does not
• Consult before you brief: let the agent interview you, then freeze the brief
• Agent briefs: scope, guardrails, checkpoints, and stop conditions
• Tools, connectors and skills: designing what an agent can reach
• Retrieval and memory patterns that survive long sessions
• Long and background runs that hand off cleanly across context windows
• Reading any AI harness on six tool-agnostic dimensions
• Calibrating an agent run versus calibrating a chat reply
• Agent security: prompt injection, enforced permissions, and never all three
• Evals for agent runs before trusting AI in production
• Subagents and orchestration patterns, and where they break
• Failure modes: over-execution, hallucinated paths, silent failures, refusal mid-loop

By Valentin Nagacevschi