# agent-camp

A Chinese-language interview-preparation knowledge base for AI agent engineering roles, organized around how interviews actually ask about LLM fundamentals, prompt and context engineering, RAG and tool use.

The knowledge base covers foundational topics such as transformers, tokenization and embeddings, pretraining and fine-tuning stages including RLHF and DPO, and inference parameters; prompt engineering techniques such as few-shot prompting, chain-of-thought and ReAct, along with prompt injection and compression; context engineering including context windows, long-context handling, memory systems and context caching; and tool use, including function-calling schema design, the Model Context Protocol, and permissions and sandboxing for custom tools.

It also covers retrieval-augmented generation techniques such as chunking, hybrid search, reranking and GraphRAG, core agent theory including ReAct, Plan-and-Execute and Reflexion, workflow orchestration tools such as LangGraph and Dify, and production engineering topics such as benchmarks, observability, harness design, cost optimization and security, alongside source-code walkthroughs of tools such as Claude Code, Codex CLI, Cline and SWE-agent.

Website: https://aigccamp.com/