Part 2: Core Capabilities
🧠 Systematically master all the core technical capabilities needed to build Agents — from tool calling to reinforcement learning training.
This is the most substantial part of the book, containing 7 chapters that cover the complete capability progression chain for Agents.
| Chapter | Content | Analogy |
|---|---|---|
| Chapter 4 Tool Calling | Function Calling, custom tools | The Agent's "hands" |
| Chapter 5 Memory System | Short-term / long-term / working memory | The Agent's "brain" |
| Chapter 6 Planning & Reasoning | ReAct, task decomposition, reflection | The Agent's "thinking" |
| Chapter 7 RAG | Document retrieval, vector databases | The Agent's "knowledge base" |
| Chapter 8 Context Engineering | Window management, attention budget | The Agent's "information hub" |
| Chapter 9 Skill System | Skill encapsulation, learning & discovery | The Agent's "skill tree" |
| Chapter 10 Agentic-RL | SFT, PPO, DPO, GRPO | The Agent's "self-evolution" |
Capability Progression: Use tools → Remember → Think → Retrieve → Manage context → Encapsulate skills → Self-evolve
Start learning: Chapter 4: Tool Calling