AI‑Assisted Code Refactoring for Go Projects: A Step‑by‑Step Guide

This article explains how to use AI (specifically the Cursor assistant) to refactor Go code efficiently. It uses the analogy of renovating an old house to illustrate the collaboration between developers and AI.

Step 1 – Scan and Analyse Code

Use a prompt such as 请分析下{待分析的代码目录}这个目录下有哪些函数过大、难以维护？请以类似 cursorrule 的格式生成列表，输出到{待分析的代码目录}/.todofunc 文件中。 to let the AI quickly identify large or hard‑to‑maintain functions and write the results to a .todofunc file. For very large codebases the scan can be split into sub‑directories with a similar prompt.

Step 2 – Discuss Refactoring Plans

Read the .todofunc file, select functions to refactor, and discuss concrete strategies with the AI (design patterns, function decomposition, etc.). Example interaction prompt: @commit 请分析{重构入口函数}及其调用的相关函数在{文件目录/代码范围}内的逻辑是否正确，是否存在问题，尤其关注：- 接口兼容性是否被破坏 - 隐式依赖是否发生变化 - 是否存在未预期的副作用 - 核心逻辑是否存在回归风险 .

Step 3 – Generate Specific Refactoring Rules

Convert the discussion outcome into structured cursorrule entries and append them to the .todofunc file, e.g., 请根据刚才讨论的方案，为该函数生成 cursorrule 规则，追加到.todofunc文件中。 .

Step 3.1 – Deep Dive into Complex Business Logic (Optional)

If a function contains intricate business logic, ask the AI to analyse it in depth and add the extracted flow to .todofunc , e.g., 请仔细分析该函数的业务逻辑，整理出核心流程并追加到.todofunc文件中。 .

Step 4 – Execute the Refactoring

Iteratively apply the generated rules, letting the AI rewrite code, add clear comments, and produce unit tests (TDD style). Prompt example (simplified):

{具体重构要求}
额外要求：1. 为重构后的函数添加详细注释。2. 不要进行其他额外操作。3. 若该函数可以编写单元测试，请同时生成单元测试代码，使用 TDD 风格，涵盖基本、特殊和边界场景。

Human reviewers must verify logic correctness, test coverage, and that no business functionality is broken.

Step 5 – AI Self‑Review

Run the AI’s @commit feature to compare before‑and‑after code, detect interface incompatibilities, hidden dependencies, side‑effects, or regression risks, and obtain suggested fixes.

Step 6 – Update Technical Documentation

Ask the AI to generate a high‑level flow diagram and a concise description of the refactored logic, then add it to project documentation, e.g., 请以 markdown 格式给出{重构入口函数}的主要流程图，不必细化到具体函数。 .

Step 7 – Full Functional Verification

Deploy the refactored code to a test environment, run end‑to‑end tests, and confirm that business requirements are still met. Human engineers perform the final validation and performance checks.

Key Collaboration Points

AI excels at fast static analysis, generating structured refactoring suggestions, and producing boilerplate (tests, docs).

Human developers provide deep domain knowledge, decide on design patterns, evaluate trade‑offs, and ensure overall system integrity.

Practical Tips

Separate responsibilities: let AI handle repetitive tasks, keep humans in charge of business decisions.

Proceed in small increments; verify after each change.

Give clear, concise prompts to the AI.

Use AI‑generated suggestions as learning material.

The guide concludes that combining AI’s speed with human judgment yields higher refactoring efficiency, better code quality, and reduced technical debt.