<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0">
  <channel>
    <title>文章 RSS</title>
    <description>编辑部文章更新。</description>
    <link>https://img.leeguoo.com/zh/archives/</link>
    <atom:link href="https://img.leeguoo.com/zh/rss.xml" rel="self" type="application/rss+xml" xmlns:atom="http://www.w3.org/2005/Atom" />
    <language>zh-cn</language>
    <lastBuildDate>Fri, 05 Jun 2026 20:40:53 GMT</lastBuildDate>
  <item>
    <title>PayPay 证券复盘 · 2026-06-05</title>
    <description>家庭账户证券复盘(业绩/成本/归因/结构/行为/昨晚),仅事实数据,非投资建议。</description>
    <link>https://img.leeguoo.com/zh/posts/paypay-review-2026-06-05/</link>
    <guid>https://img.leeguoo.com/zh/posts/paypay-review-2026-06-05/</guid>
    <category>Editorial</category>
    <pubDate>Fri, 05 Jun 2026 09:15:42 GMT</pubDate>
  </item>
  <item>
    <title>Claude Code Workflows 深度解析：把编排写进脚本之后</title>
    <description>Claude Code 昨天（2026-05-28）放出 dynamic workflows，目前还是 research preview。这篇不讲源码，只讲它到底改了什么、什么时候用划算、什么时候纯属浪费 token——以及它和 subagent / skill 的本质区别。</description>
    <link>https://img.leeguoo.com/zh/posts/claude-code-workflows-deep-dive/</link>
    <guid>https://img.leeguoo.com/zh/posts/claude-code-workflows-deep-dive/</guid>
    <category>Editorial</category>
    <pubDate>Fri, 29 May 2026 09:38:38 GMT</pubDate>
  </item>
  <item>
    <title>状态栏那行 cache 4m23s，到底准不准？</title>
    <description>状态栏上那行 prompt-cache 倒计时是怎么算出来的，准不准，会不会骗你。</description>
    <link>https://img.leeguoo.com/zh/posts/claude-statusbar-cache-countdown/</link>
    <guid>https://img.leeguoo.com/zh/posts/claude-statusbar-cache-countdown/</guid>
    <category>Editorial</category>
    <pubDate>Fri, 29 May 2026 04:55:36 GMT</pubDate>
  </item>
  <item>
    <title>东京通勤圈购房候选清单</title>
    <description>面向长期自住、通勤、防灾和预算压力的购房候选对比页</description>
    <link>https://img.leeguoo.com/zh/posts/tokyo-house-candidates/</link>
    <guid>https://img.leeguoo.com/zh/posts/tokyo-house-candidates/</guid>
    <category>Editorial</category>
    <pubDate>Wed, 27 May 2026 08:57:48 GMT</pubDate>
  </item>
  <item>
    <title>从软方法到硬补丁 — 一次 macOS Mach-O 逆向的方法论复盘</title>
    <description>一次 8 小时 macOS 二进制逆向工程的方法论笔记 — 软方法为什么会失败、什么时候必须切到 hex patch、Ghidra/lldb/llvm-objdump 怎么配合、以及&quot;杀手补丁&quot;这种思路在攻防演练里能怎么用。不公开具体 target，只讲方法、工具与排错技巧。</description>
    <link>https://img.leeguoo.com/zh/posts/macos-binary-rev-methodology/</link>
    <guid>https://img.leeguoo.com/zh/posts/macos-binary-rev-methodology/</guid>
    <category>Editorial</category>
    <pubDate>Tue, 26 May 2026 17:51:19 GMT</pubDate>
  </item>
  <item>
    <title>ChatGPT 订阅 → 生图 API 可视化速览：chatgpt-imagegen 一图看懂</title>
    <description>一页可视化速览：ChatGPT 订阅是怎么被 chatgpt-imagegen 转成命令行生图工具的——OAuth 取 token、Responses API SSE 流、能力边界矩阵、不该用的场景一图说清。</description>
    <link>https://img.leeguoo.com/zh/posts/chatgpt-imagegen-visual-guide/</link>
    <guid>https://img.leeguoo.com/zh/posts/chatgpt-imagegen-visual-guide/</guid>
    <category>Editorial</category>
    <pubDate>Tue, 26 May 2026 07:54:43 GMT</pubDate>
  </item>
  <item>
    <title>直面 AI 审计：如何将逆向工程项目&quot;合法化&quot;？</title>
    <description>探讨一种结合 Claude Code 与 Gemini CLI 来处理反编译项目的实践方法。聚焦于所有权验证这一难题，并介绍通过自动清理处理来解决该问题的思路。</description>
    <link>https://img.leeguoo.com/zh/posts/ai-assisted-sanitization-workflow/</link>
    <guid>https://img.leeguoo.com/zh/posts/ai-assisted-sanitization-workflow/</guid>
    <category>Editorial</category>
    <pubDate>Tue, 26 May 2026 07:42:15 GMT</pubDate>
  </item>
  <item>
    <title>日语鸿蒙 v8 Bug 表</title>
    <description>v8 122 条 bug 修复 — 真装机验证进行中（12 条已实证 + 2 条因模拟器问题待重测；剩余 108 条排队等待用户配合解锁继续）</description>
    <link>https://img.leeguoo.com/zh/posts/japanese-harmany-bug-v8-skeleton/</link>
    <guid>https://img.leeguoo.com/zh/posts/japanese-harmany-bug-v8-skeleton/</guid>
    <category>Editorial</category>
    <pubDate>Wed, 13 May 2026 03:02:22 GMT</pubDate>
  </item>
  <item>
    <title>HTML Raw 模式样例</title>
    <description>测试 raw 模式：用户 HTML 替换整页，自由设计。</description>
    <link>https://img.leeguoo.com/zh/posts/html-raw-demo-2026/</link>
    <guid>https://img.leeguoo.com/zh/posts/html-raw-demo-2026/</guid>
    <category>Editorial</category>
    <pubDate>Tue, 12 May 2026 10:33:36 GMT</pubDate>
  </item>
  <item>
    <title>HTML 主题模式样例</title>
    <description>测试 themed 模式：CSS 自动 scope，保留博客排版，SEO 友好。</description>
    <link>https://img.leeguoo.com/zh/posts/html-themed-demo-2026/</link>
    <guid>https://img.leeguoo.com/zh/posts/html-themed-demo-2026/</guid>
    <category>Editorial</category>
    <pubDate>Tue, 12 May 2026 10:33:30 GMT</pubDate>
  </item>
  <item>
    <title>/goal 命令实现剖析</title>
    <description>基于 Claude Code 2.1.139 headless 版源码，拆解 /goal 命令如何通过会话级 Stop hook 驱动 Agent 持续工作直到条件满足。</description>
    <link>https://img.leeguoo.com/zh/posts/goal-command-implementation/</link>
    <guid>https://img.leeguoo.com/zh/posts/goal-command-implementation/</guid>
    <category>Editorial</category>
    <pubDate>Tue, 12 May 2026 01:40:30 GMT</pubDate>
  </item>
  <item>
    <title>claude agents 与 /bg：Agent View 实现剖析</title>
    <description>基于 Claude Code 2.1.139 源码，拆解 claude agents、--bg 与 /bg 如何共用后台 job 模型、daemon 与 FleetView TUI。</description>
    <link>https://img.leeguoo.com/zh/posts/claude-agents-bg-implementation/</link>
    <guid>https://img.leeguoo.com/zh/posts/claude-agents-bg-implementation/</guid>
    <category>Editorial</category>
    <pubDate>Tue, 12 May 2026 01:40:30 GMT</pubDate>
  </item>
  <item>
    <title>为浏览器自动化设计攻防策略：检测模型与分层控制平面</title>
    <description>将高度对抗环境中的浏览器自动化抽象为多维风险评分系统，并基于一致性、稀有性和时间分布三个核心维度构建分层控制平面。</description>
    <link>https://img.leeguoo.com/zh/posts/agent-browser-stealth-attack-defense-playbook/</link>
    <guid>https://img.leeguoo.com/zh/posts/agent-browser-stealth-attack-defense-playbook/</guid>
    <category>Editorial</category>
    <pubDate>Tue, 28 Apr 2026 01:41:38 GMT</pubDate>
  </item>
  <item>
    <title>我是如何确认 WeChat macOS 4.0.1.52 本地数据库解锁链路的</title>
    <description>一篇以本人设备为对象的工程取证复盘。文章按顺序追踪了静态文件、运行时打开路径以及 SQLCipher 参数，最终确认 WeChat macOS 4.0.1.52 的本地数据库解锁链路，是建立在账号级密钥分发模型之上的。</description>
    <link>https://img.leeguoo.com/zh/posts/wechat-macos-db-runtime-analysis-20260323/</link>
    <guid>https://img.leeguoo.com/zh/posts/wechat-macos-db-runtime-analysis-20260323/</guid>
    <category>Editorial</category>
    <pubDate>Mon, 23 Mar 2026 06:40:18 GMT</pubDate>
  </item>
  <item>
    <title>做自己的微信账号自动化，应该先想清楚哪三层</title>
    <description>如果目标是“自己的账号自动化”，真正该拆开的不是功能清单，而是监听层、执行层和存储层。</description>
    <link>https://img.leeguoo.com/zh/posts/wechat-personal-account-automation-strategy-20260319/</link>
    <guid>https://img.leeguoo.com/zh/posts/wechat-personal-account-automation-strategy-20260319/</guid>
    <category>Editorial</category>
    <pubDate>Thu, 19 Mar 2026 08:26:37 GMT</pubDate>
  </item>
  <item>
    <title>从辅助编码到一体化研发：AI 编程团队落地实践（PPT）</title>
    <description>一份 Slidev 风格的中文演示稿：解释 Agentic Coding、Skill 与 MCP、AGENTS.md 与 CLAUDE.md、沙箱和权限控制，以及团队如何用 Plan、YApi Skill 和 docs-sync 把 AI 编程真正落地。</description>
    <link>https://img.leeguoo.com/zh/posts/ai-coding-agentic-coding-team-practice-ppt-20260318/</link>
    <guid>https://img.leeguoo.com/zh/posts/ai-coding-agentic-coding-team-practice-ppt-20260318/</guid>
    <category>Editorial</category>
    <pubDate>Wed, 18 Mar 2026 08:25:15 GMT</pubDate>
  </item>
  <item>
    <title>从辅助编码到 Agentic Coding：团队如何真正把 AI 编程用起来</title>
    <description>围绕 Agentic Coding 的核心问题，系统梳理主流工具形态、Skill 与 MCP 的边界、AGENTS.md 与 CLAUDE.md 的作用原理、沙箱与权限控制，以及如何用真实任务和 benchmark 评估 AI 编程，而不是只看 demo。</description>
    <link>https://img.leeguoo.com/zh/posts/agentic-coding-skill-mcp-agents-md-claude-md-principles-20260318/</link>
    <guid>https://img.leeguoo.com/zh/posts/agentic-coding-skill-mcp-agents-md-claude-md-principles-20260318/</guid>
    <category>Editorial</category>
    <pubDate>Wed, 18 Mar 2026 05:28:12 GMT</pubDate>
  </item>
  <item>
    <title>QClaw 的实现原理：它如何把 OpenClaw 产品化成桌面应用</title>
    <description>这篇文章不再停留在“控制面 / 执行面”的总论，而是基于当前实现把桥接层、IPC、配置字段、微信链路、回滚机制和证据索引一一展开，说明 QClaw 如何把 OpenClaw 组织成一个可交付的桌面运行时。</description>
    <link>https://img.leeguoo.com/zh/posts/qclaw-implementation-principles-20260310/</link>
    <guid>https://img.leeguoo.com/zh/posts/qclaw-implementation-principles-20260310/</guid>
    <category>Editorial</category>
    <pubDate>Tue, 10 Mar 2026 08:25:26 GMT</pubDate>
  </item>
  <item>
    <title>OpenClaw v2026.3.8 发布解读：备份、远程网关、Talk 模式与多端路由继续补强</title>
    <description>基于官方 release notes，总结 OpenClaw v2026.3.8 的关键变化：备份命令落地、macOS 远程网关 onboarding、Talk silence timeout、Brave 搜索接入、ACP receipts，以及多平台路由修复。</description>
    <link>https://img.leeguoo.com/zh/posts/openclaw-v2026-3-8-release-analysis-20260309/</link>
    <guid>https://img.leeguoo.com/zh/posts/openclaw-v2026-3-8-release-analysis-20260309/</guid>
    <category>Editorial</category>
    <pubDate>Mon, 09 Mar 2026 08:32:52 GMT</pubDate>
  </item>
  <item>
    <title>设计浏览器自动化攻防：检测模型与分层控制平面</title>
    <description>本文将高对抗环境中的浏览器自动化抽象为一个多维风险评分系统，并围绕一致性、稀有性与时间分布三大核心维度构建分层控制平面。</description>
    <link>https://img.leeguoo.com/zh/posts/agent-browser-stealth-defense-playbook-v2/</link>
    <guid>https://img.leeguoo.com/zh/posts/agent-browser-stealth-defense-playbook-v2/</guid>
    <category>Editorial</category>
    <pubDate>Fri, 06 Mar 2026 10:16:32 GMT</pubDate>
  </item>
  <item>
    <title>Cloudflare Turnstile 攻防方案设计：系统原理与控制面</title>
    <description>从能力令牌视角重构 Turnstile 风险模型，聚焦签发/消费语义、作用域绑定和执行完整性，给出高对抗场景下的防护优先级。</description>
    <link>https://img.leeguoo.com/zh/posts/cloudflare-turnstile-stability-principles/</link>
    <guid>https://img.leeguoo.com/zh/posts/cloudflare-turnstile-stability-principles/</guid>
    <category>Editorial</category>
    <pubDate>Tue, 03 Mar 2026 15:06:27 GMT</pubDate>
  </item>
  <item>
    <title>使用 agent-browser-stealth 代替 agent-browser</title>
    <description>面向推广场景：在亚马逊等高风控站点提升 AI 浏览器可操作性，并支持复用用户浏览器状态。</description>
    <link>https://img.leeguoo.com/zh/posts/use-agent-browser-stealth-instead-of-agent-browser/</link>
    <guid>https://img.leeguoo.com/zh/posts/use-agent-browser-stealth-instead-of-agent-browser/</guid>
    <category>Editorial</category>
    <pubDate>Tue, 03 Mar 2026 03:29:58 GMT</pubDate>
  </item>
  </channel>
</rss>