AgentGuide
A Chinese-language, systematic learning guide for AI agent engineering and research, modeled on JavaGuide, covering agent systems, context engineering, RAG, evaluation and reinforcement learning post-training, oriented toward interview preparation.
AgentGuide is a Chinese-language open knowledge base for AI agent engineering, research and job hunting, modeled on the structure of JavaGuide. It curates existing courses, tutorials and papers rather than duplicating them, and organizes them into learning paths for either an engineering-focused role or an algorithm-focused role.
Its content covers agent systems and harnesses (the agent loop, workflow versus agent versus multi-agent boundaries, state, scheduling and permissions), context engineering and memory, tools and protocols such as MCP, Skills and A2A/ACP, retrieval-augmented generation including GraphRAG and agentic RAG, evaluation and observability, and post-training methods such as SFT, DPO, GRPO and reinforcement learning. It also includes an interview question bank of more than 1,500 questions, system-design material, and guidance on writing a resume and portfolio project for these roles.
It links to a related project, learn-workbuddy, a clean-room recreation of a desktop agent harness covering the agent loop, tool calling, context engineering, long-term memory and permission auditing.