回顧與練習
關鍵要點
- 上線是一套工作流,而不是一個瞬間:部署、觀察、修補、重複 — 並且把回滾做成你已經測試過一次的東西。
- 在宣佈任何東西之前先過一遍上線前檢查清單,讓 AI 逐項核驗,而不要依賴你的記憶。
- 讓測量來驅動最佳化 — Lighthouse 報告永遠勝過"感覺很慢"的直覺。
- 金鑰放在生產配置裡,絕不放進倉庫;
.env加入 gitignore,.env.example保持最新。 - 第一個版本是假設,真實使用者才是實驗,所以要用小而快的修復來迭代。
動手試試
挑一個你已部署的應用(或現在就部署一個很小的),逐行對照走一遍完整的上線前檢查清單。對每一項你無法誠實勾選的,把失敗的那一項貼給 AI,請它給出彌補差距的最小改動。最後點選線上站點的一個深層連結 — 而不只是首頁 — 確認生產環境的行為和開發環境一致。
本章提示詞
Here is my deployed app and its stack: [describe app + hosting + services].
Act as my launch reviewer. Walk this pre-launch checklist item by item
and, for each, tell me how to verify it on the LIVE site (not in dev):
- production URL loads over valid HTTPS on the custom domain
- all secrets set in production; none committed to the repo
- core happy path (sign-up, login, main action) works in production
- payments tested end-to-end including the webhook (if any)
- error tracking and uptime monitoring are live
- I can roll back, and have tested it once
- mobile layout works on a real phone
- database backups are enabled
For every item that fails, give me the SMALLEST change that fixes it.
Do not assume anything passes — make me prove each one.