The Agentic Handbook

6 Lessons

The deeper track: where Tactical Agentic Coding is CLI-first, the Agentic Handbook goes deep on context, prompts, and multi-agent systems, mostly built programmatically on the Claude Agent SDK.

If Tactical Agentic Coding teaches you the tactics, the Agentic Handbook teaches you how to build the machinery: how to engineer context so a focused agent performs at its peak, how to write prompts that fire hundreds of tool calls correctly, how to build custom agents from scratch, how to orchestrate a fleet of them, and how to make them learn. The final lesson ties it all together into the codebase singularity.

The six lessons

#	Lesson	Core idea
1	Elite Context Engineering	The R&D framework (Reduce, Delegate) and 12 context techniques across 4 levels.
2	Agentic Prompt Engineering	The Stakeholder Trifecta and the Seven Levels of agentic prompt formats.
3	Building Domain-Specific Agents	Better, then more, then custom agents, built on the Claude Agent SDK.
4	The Orchestrator Agent	One interface commanding a fleet of primary agents. Agents that conduct agents.
5	Agent Experts	Self-improving agents. Act, learn, reuse. Expertise as a maintained mental model.
6	Codebase Singularity	The classes-and-grades maturity model, and the capstone: your codebase runs itself.

The progression

flowchart LR
  A[Base agent] --> B[Better agent]
  B --> C[More agents]
  C --> D[Custom agents]
  D --> E[Orchestrator]
  E --> F[Agent experts]
  F --> G((Codebase
Singularity))

The through-line: everything becomes the Core Four. Whether you build a prompt, a sub-agent, a skill, or a fully custom agent with its own system prompt and tools, you are always configuring context, model, prompt, and tools. The composite type does not matter; the leverage does.