walkinglabs/learn-harness-engineering：README 來源編輯指南

walkinglabs/learn-harness-engineering 的 README 將專案描述為「Harness engineering beginner tutorial, from 0 to 1」。本文只整理倉庫可直接核對的內容，不把 star、Fork 或宣傳語當成品質證明。README 在「README」下寫到：> 🌍 This course is available in 15 languages: English, 简体中文, 繁體中文, 日本語, 한국어, Español, Français, Русский, Deutsch, العربية, Tiếng Việt, Oʻzbekcha, Türkçe, Portuguese (BR), Українська. Choose your language from the badges above.。這說明的是專案邊界，不是已完成的生產驗證。

從 README 的「Core References」與相關條目，可以先判斷它是否處理你的實際問題：Anthropic: Effective harnesses for long-running agents。若需求不同，不應只因專案熱度就採用。本文保留原始專案名、命令與元件名，方便回到一手來源核對。 README 另外列出一項可核對的資訊：OpenAI: Harness engineering: leveraging Codex in an agent-first world。這類原文條目可用來設計試跑步驟，但不能取代實際環境測試。

README 將運作方式分散在「🆕 What's New , August 2026」等段落。可確認的線索包括：| What | Details | |------|---------| | Lecture 14 | From Single Loops to Graph Engineering , Three progressive experiments: draw your maker-checker loop as an explicit graph, add a parallel fan-out/fan-in node, then add a conditional。本文不把未寫出的架構、效能或安全邊界補成結論；真正的執行鏈仍要配合目錄、設定檔與版本標籤檢查。

第一次安裝應從 README 指出的入口開始。目前可核對的命令是： THE HARNESS PATTERN ==================== You --> give task --> Agent reads harness files --> Agent executes | harness governs every step: | +--> Instructions: what to do, in what order +--> Scope: one feature at a time, no overr 如果倉庫沒有命令，本文不會自行編造步驟，而是建議先閱讀「🆕 What's New , July 2026」，確認系統依賴、預設埠與首次初始化。

日常使用取決於專案文件。README 的「🆕 What's New , August 2026」段落提到：Key idea: A loop is a graph with one node. When your task needs specialization, parallelism, shared state, verification, and recovery , it has stopped being a loop. It's a graph.。設定檔、環境變數、權限與資料目錄只在來源明確時才會記錄；沒有寫出的預設值，應在測試環境驗證並保留回滾副本。 同一部分也提到：Anthropic: Harness design for long-running application development。