📄 中文摘要
AI 代理在编辑代码时，如果没有合适的工具，可能会导致严重问题。例如，当代理被要求为登录函数添加错误处理时，可能会生成 200 行代码并引入语法错误，导致构建失败。标准工具会将这些错误代码写入磁盘，导致持续集成（CI）失败，开发者需要花费时间进行调试。代码外科刀通过语法树（AST）解析器在代码写入磁盘之前捕捉这些错误，从而提高开发效率，避免不必要的时间浪费。
代码外科刀：让软件开发变得更便宜、可管理、准确和安全的 AI 代理
📄 English Summary
Code Scalpel: Making Software Development with AI Agents Cheaper, Governable, Accurate, and Safe
When an AI agent edits code without the right tools, it can lead to significant issues. For instance, when tasked with adding error handling to a login function, the agent might regenerate 200 lines of code and introduce a syntax error, causing the build to fail. Standard tools write this erroneous code to disk, leading to continuous integration (CI) failures and requiring developers to spend time debugging. Code Scalpel addresses this by using an Abstract Syntax Tree (AST) parser to catch errors before they touch the disk, thereby enhancing development efficiency and preventing unnecessary time wastage.