编码Agent主要读指令文件而非API文档，AGENTS.md应更受重视
Coding agents mostly read the instruction files and notes they were given or wro…
Coding agents mostly read the instruction files and notes they were given or wrote themselves, so your AGENTS.md deserves more care than your API reference.
编码代理主要阅读它们被给予或自己编写的指令文件和笔记，因此你的AGENTS.md比你的API参考更值得关注。
because that is what agents actually read, not your API reference.
因为那是代理实际阅读的内容，而不是你的API参考。
The authors tracked 3,033 documentation events across 557 real agent coding sessions, plus file changes in 33,097 agent-written pull requests.
作者追踪了557次真实代理编码会话中的3,033个文档事件，以及33,097个代理编写的拉取请求中的文件变更。
Instruction files like AGENTS.md and CLAUDE.md were 35.4% of that activity, and the agent's own plans and notes another 25.1%. API references got 1.3%, troubleshooting guides 0.4%. Agents also wrote these files nearly as often as they read them.
像AGENTS.md和CLAUDE.md这样的指令文件占这些活动的35.4%，代理自己的计划和笔记又占25.1%。API参考仅占1.3%，故障排除指南占0.4%。代理编写这些文件的频率几乎与阅读它们的频率相当。
More importantly, the assumed read-the-docs-then-write-the-code loop barely showed up. Only 3 of 1,328 documentation reads led straight to a code edit, and the authors found no case of an agent checking its work against documentation.
更重要的是，假定的先读文档再写代码的循环几乎没有出现。在1,328次文档阅读中，只有3次直接导致了代码编辑，而且作者没有发现任何代理根据文档检查其工作的案例。
– arxiv. org/abs/2608.20195
– arxiv.org/abs/2608.20195
Title: "From Agent Behaviour to Agent-Friendly Documentation: An Empirical Study of How Coding Agents Discover, Read, and Write Technical Documentation"
标题：“从代理行为到代理友好文档：编码代理如何发现、阅读和编写技术文档的实证研究”
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力