<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:sy="http://purl.org/rss/1.0/modules/syndication/" xmlns:media="http://search.yahoo.com/mrss/"><channel><title>Codex on k4i's blog</title><link>https://k4i.top/zh/tags/codex/</link><description>Recent content in Codex on k4i's blog</description><generator>Hugo -- gohugo.io</generator><language>zh</language><managingEditor>sky_io@outlook.com (K4i)</managingEditor><webMaster>sky_io@outlook.com (K4i)</webMaster><copyright>All content is subject to the license of &lt;a rel="license noopener" href="https://creativecommons.org/licenses/by-nc-sa/4.0/" target="_blank"&gt;CC BY-NC-SA 4.0&lt;/a&gt; .</copyright><lastBuildDate>Sat, 23 May 2026 10:30:00 +0800</lastBuildDate><atom:link href="https://k4i.top/zh/tags/codex/index.xml" rel="self" type="application/rss+xml"/><item><title>Agent Skill Management：把 AI 助手从聪明变成稳定</title><link>https://k4i.top/zh/posts/agent-skill-management/</link><pubDate>Sat, 23 May 2026 10:30:00 +0800</pubDate><author>sky_io@outlook.com (K4i)</author><atom:modified>Sat, 30 May 2026 00:36:50 +0800</atom:modified><guid>https://k4i.top/zh/posts/agent-skill-management/</guid><description>&lt;h2 id="why-skills-matter"&gt;为什么需要 Skill&lt;/h2&gt;
&lt;p&gt;用 coding agent 一段时间后，我发现真正影响效率的不是“模型今天够不够聪明”，而是它能不能稳定地复用经验。&lt;/p&gt;
&lt;p&gt;比如同一个仓库里，测试命令是什么、哪些文件不能随便改、生成图片应该放在哪个目录、发布前要跑什么检查、遇到 bug 应该先定位根因还是先补丁。人可以靠记忆和习惯处理这些隐性流程，但 agent 每次启动时并不天然知道这些约定。把所有规则都塞进一份巨大的 &lt;code&gt;AGENTS.md&lt;/code&gt; 也不是好办法：上下文会膨胀，触发边界会模糊，最后变成“什么都写了，但关键时刻不一定用”。&lt;/p&gt;</description><dc:creator>K4i</dc:creator><media:content url="https://k4i.top//images/icons/terminal.png" medium="image"><media:title type="html">featured image</media:title></media:content><category>agent</category><category>codex</category><category>skills</category><category>workflow</category><category>notes</category></item></channel></rss>