Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

Story

Jin links AGENTS.md rules to coding workflow stages

AI summary

Jin’s author says the coding agent reads two layers of rules on every invocation: repository-root AGENTS.md files for shared project constraints and global files for personal preferences. These rules are injected into semantic scanning, planning, and implementation, including method-level intent checks, while files exceeding 16,000 characters are truncated. The report provides no independent test of rule compliance or update timing.

In-depth insight

Jin moves repository rules beyond conversational prompts into the stages that produce architecture and code, giving coding-agent users a more structured way to express project and personal constraints.

Jin 开源地址:github.com/ZYF93/Jin | 三平台安装包见 Releases

现在很多 coding agent 都支持 "rules 文件":写一个 AGENTS.md,放几条项目约定,模型"会尽量遵守"。用久了你会发现一个尴尬的事实——规则到底在哪一步生效、生效到什么粒度,是不确定的。

Jin 里我把这件事做成了确定链路:两层规则文件,注入三个环节,一路管到方法级。

两层文件,各管各的

层位置适合放什么
项目规则仓库根目录 AGENTS.md分层约束、命名、测试要求、领域知识(跟着仓库走,队友共享)
全局规则应用数据目录(设置里直接编辑)个人偏好:语言、回复风格、常用习惯(跟机器走,不进仓库)

每次调用实时读取——改完文件,下一次扫描 / 规划 / 实现立即用新规则,不用重启。单层超过 16,000 字符会截断,防止规则把每个请求撑爆。

#Jin#AGENTS.md#LSP

Original · 3 min readRead the original
Group chat

No comments yet. Start the conversation.