📄 中文摘要
编码代理的定义特征在于它能够执行自己编写的代码，这使得编码代理比仅仅输出代码而无法验证的语言模型（LLMs）更为有用。生成的代码在未执行之前，不能假设其有效性。编码代理能够确认其生成的代码是否按预期工作，或者在必要时进一步迭代代码以确保其功能的正确性。让代理编写单元测试，尤其是采用测试驱动开发（TDD）的方法，是确保其代码有效性的重要手段。
自主手动测试
发布: 2026年3月6日
📄 English Summary
Agentic manual testing
The defining characteristic of a coding agent is its ability to execute the code it writes, making coding agents significantly more useful than language models (LLMs) that generate code without verification. Code generated by an LLM should not be assumed to work until it has been executed. Coding agents can confirm that the code they produce works as intended or iterate on it until it does. Encouraging agents to write unit tests, particularly using test-driven development (TDD), is a powerful way to ensure the effectiveness of their code.