AI Auto-Work:定義 AI 驅動的全週期軟體工程自動化 / AI Auto-Work: Defining the AI-Driven Full-Cycle Software Engineering Automation

從需求研究到代碼提交的端到端自動化開發工作流 / An end-to-end automated development workflow from requirement research to code commit.

🔎 工具速覽 / AT A GLANCE

CategoryAI Agentic Coding Workflow
PricingOpen Source (MIT License)
BestForMedium to Large-scale Engineering Projects
GitHub Stars⭐ 89

🚀 引言 / Introduction

AI Auto-Work 是一個革命性的代理編碼工作流系統,它將 LLM 的能力從單純的代碼補全提升到了完整的工程管理層級。透過 Claude 與 Codex 的雙模型協作,實現了從需求分析、規劃到開發與審查的閉環自動化。

AI Auto-Work is a revolutionary agentic coding workflow system that elevates LLM capabilities from simple code completion to full-scale engineering management, utilizing a dual-model collaboration between Claude and Codex to achieve a closed-loop automation of research, planning, development, and review.

🛠️ 核心功能 / Key Features

Full-Cycle Autonomous Development: Covers the complete lifecycle from Requirement Research → Review → Planning → Review → Development → Review.

全週期自主開發:涵蓋需求研究 → 審查 → 規劃 → 審查 → 開發 → 審查的完整生命週期。

Cross-Model Adversarial Review: Claude acts as the executor while Codex performs adversarial auditing to eliminate single-model blind spots.

跨模型對抗審查:由 Claude 執行開發,Codex 負責對抗性審查,有效消除單一模型的盲點。

Zero Context Pollution: Each stage runs in an isolated process with handoffs via persisted files only, ensuring logical purity.

零上下文污染:每個階段在獨立進程中運行,僅通過持久化文件傳遞,確保邏輯純淨。

Mechanical Quality Gates: Mandatory compile and test passing requirements before any review cycle begins to prevent trivial errors.

機械化質量門禁:強制要求編譯與測試通過後方可進入審查環節,杜絕低級錯誤。

Atomic Commits: Adheres to the 'one task, one commit' principle, ensuring code is fully bisectable and traceable.

原子化提交:遵循「一個任務,一次提交」原則,確保代碼可追溯且易於回溯。

💡 技術亮點 / Tech Highlights

Triple-Check Convergence Loop: Delivers production-quality code through a rigorous loop of execution, testing, and review.

三重檢查收斂迴路:通過執行-測試-審查的循環,交付生產級別的高質量代碼。

Self-Improving Context: Systematic errors update a shared knowledge base (.ai/) rather than just modifying the code.

自進化上下文:系統性錯誤會自動更新至共享知識庫 (.ai/),而非僅修改代碼。

Flexible Control Modes: Supports fully automated, semi-automated (human checkpoints), and independent module calls.

靈活的控制模式:支持全自動、半自動(人工檢查點)以及單一功能模組的獨立調用。

📦 快速上手 / Quick Start

Fully Automated Workflow: /auto-work [feature description] — One-click completion from requirement to commit.

全自動工作流:/auto-work [功能描述] — 從需求到提交的一鍵式完成。

Fast Path: /fast-auto-work [version] [task] [description] — Rapid path for small direct changes.

快速修正:/fast-auto-work [版本] [任務] [描述] — 針對小規模變更的快速路徑。

Manual Mode: /manual-work [module] — Incorporates human checkpoints at key milestones.

人工干預模式:/manual-work [模組名] — 在關鍵里程碑加入人類審查。

Tech Research: /research:do [topic] — Deep analysis for technology selection.

技術研究:/research:do [研究主題] — 針對技術選型進行深入分析。

準備好試試 AI Auto-Work:定義 AI 驅動的全週期軟體工程自動化 / AI Auto-Work: Defining the AI-Driven Full-Cycle Software Engineering Automation 了嗎?

Ready to try AI Auto-Work:定義 AI 驅動的全週期軟體工程自動化 / AI Auto-Work: Defining the AI-Driven Full-Cycle Software Engineering Automation?

前往 GitHub 頁面 →

留言

這個網誌中的熱門文章

[Security] wpa_supplicant setup

[拆機] Nexus5 更換背蓋、電池

[我的MAC Air] 2012年中,MAC Air SSD升級