Description
随着越来越多的团队深入 Agent 工程开发，上下文管理成为无法回避的核心挑战。本视频探讨大规模 Agent 上下文管理的关键概念：Agent 如何访问、存储信息，以及如何将每次运行的经验沉淀为未来 Agent 可用的知识。视频还分析了 Claude.md 文件膨胀的问题，并介绍更成熟、规范化的上下文存储方式，帮助你的 Agent 在处理代码库时更加智能高效。
As agentic engineering scales across teams, context management becomes a critical challenge. This video covers how agents access and store information, how learnings are distilled across runs, and how to move beyond bloated Claude.md files toward a more mature, formalized approach to context storage — so your agents are better informed every time they work on your codebase.