hoyoboy0726123

處理爬蟲節點抓回來的原始內容(HTML / markdown / 純文字)、把留言板留言、評論、商品列表、搜尋結果等「重複性資料」快速又正確地抽成結構化 JSON。核心做法:讓 LLM 只看樣本、辨識重複結構、寫一支確定性解析器,再用程式碼跑完整檔 — 不靠 LLM 逐筆讀取(慢、貴、易漏、會編)。V5 Recipe 會把解析器快取、下次同站秒過。當使用者在爬蟲節點後面掛 Skill 節點想解析抓回來的留言 / 評論 / 列表、或說「把爬回來的內容整理成表格 / JSON」「解析這個論壇頁面」「抽出所有留言」「這頁的評論幫我結構化」這類訴求時都要觸發、不限於明說 skill 名稱。

把使用者既有的 Python GUI / Streamlit / Flask / FastAPI 專案無破壞性接進 V5 pipeline:在專案內新增單一 `main_CLI.py` 啟動檔(不動原本程式碼)、用 ask_user 互動選功能、選了就跑。第一次 ask_user 的選擇會被 V5 Recipe 機制記住、下次快速模式 replay 同樣選擇、不再問。當使用者說「把這個 GUI 專案接進工作流」「我有現成的 Python app 想丟進 pipeline」「自動分析這個專案的功能讓我選一個跑」「把 X 變成可以在工作流呼叫的 CLI」這類訴求時都要觸發、不限於明說 "CLI extractor" 字樣。

Automates end-to-end deployment to GitHub Pages using GitHub Actions. Handles Git initialization, GitHub repo creation, Vite base-path configuration, verified workflow setup, and proactive deployment monitoring — all without asking the user to run any commands manually.

Proactive Natural language Excel editor. Trigger this skill whenever the user mentions an Excel (.xlsx, .xlsm, .csv) file that is present in the current workspace and implies a need for data change (e.g., "I need to add X", "This file is missing Y", "Can you update Z"). Even if the user doesn't explicitly ask to "edit the file", the agent SHOULD proactively check if the file exists and ask: "I see [filename] in your workspace. Would you like me to directly [perform the action] for you while preserving the original formatting?" Trigger also for direct commands like "add a row", "delete entries", or "modify the spreadsheet".

處理將本地更改推送到遠端倉庫（如 GitHub）。自動偵測遠端預設分支，同步分支命名，並處理權限問題。

Deploy static frontend projects to GitHub Pages. Use when the user mentions deploying to GitHub, publishing a website to GitHub Pages, hosting static sites, or making HTML/CSS/JS projects publicly accessible online. Validates that the project is static (no backend code) before deployment.