<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>E2E on Peter.H's Full-Stack GAME</title><link>https://peter-notes.pages.dev/tags/e2e/</link><description>Recent content in E2E on Peter.H's Full-Stack GAME</description><image><title>Peter.H's Full-Stack GAME</title><url>https://peter-notes.pages.dev/images/og-image.jpg</url><link>https://peter-notes.pages.dev/images/og-image.jpg</link></image><generator>Hugo</generator><language>zh-tw</language><lastBuildDate>Fri, 15 May 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://peter-notes.pages.dev/tags/e2e/index.xml" rel="self" type="application/rss+xml"/><item><title>LLM 驅動的 E2E 驗證：為什麼 bot 比人更容易揭發 backend regression</title><link>https://peter-notes.pages.dev/posts/llm-driven-e2e-surfaces-backend-regression/</link><pubDate>Fri, 15 May 2026 00:00:00 +0000</pubDate><guid>https://peter-notes.pages.dev/posts/llm-driven-e2e-surfaces-backend-regression/</guid><description>用 Playwright MCP 讓 LLM 驅動瀏覽器跑 STG 驗收，原本目標是確認 frontend 修補有效，卻意外揭發 backend dual-write lifecycle hook 失效的 regression。POST 200 不等於資料寫對：探討為什麼 bot 會多走那一步、人手 UAT 為什麼通常碰不到、以及如何把這個習慣寫進 verification recipe。</description></item></channel></rss>