📄 中文摘要
在处理遗留或未知代码库时，首先应通过向Cursor、Claude Code或Codex提问进行研究。确保在AI修改后进行最小测试，以验证代码中没有错误，包括代码检查和类型检查。一些代理需要特定规则，而其他代理则会进行测试。手动测试每次代码更改至关重要，因此设置本地调试工具（如脚本、命令和用户界面/浏览器交互）非常有价值。未来可能会因迭代速度更快而省略此步骤。
2026年功能编码
发布: 2026年2月17日
📄 English Summary
Coding a Feature in 2026
When dealing with a legacy or unknown codebase, it is advisable to conduct research by asking questions to tools like Cursor, Claude Code, or Codex. Following AI modifications, it is essential to perform minimal testing to ensure there are no errors in the code, including linting and type checking. Some agents may require specific rules, while others will always attempt to conduct testing. Manual testing of every code change is crucial, making the setup of local debugging tools (such as scripts, commands, and UI/browser interactions) highly valuable. This step may be skipped in the future as iterations become faster and safer.