<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>明天</title><description>把值得反复看的技术内容沉淀下来：GitHub 一周热点精选、大厂 Agent 面试题整理。只写&quot;为什么值得看&quot;，不搬运数字。</description><link>https://mingtian-ten.vercel.app/</link><language>zh-cn</language><item><title>Agent / 大模型 / Skill / MCP：GitHub 热门项目与想法纵览</title><link>https://mingtian-ten.vercel.app/c/github-weekly/agent-llm-skill-mcp-landscape/</link><guid isPermaLink="true">https://mingtian-ten.vercel.app/c/github-weekly/agent-llm-skill-mcp-landscape/</guid><description>Agent 生态的重心正从&quot;框架&quot;移向 harness 与协议。本文把四条赛道上最值得关注的开源项目摊开：各自解决哪一层的问题、量级如何、什么场景该选谁。</description><pubDate>Tue, 29 Sep 2026 11:39:56 GMT</pubDate><category>开源</category><category>工具</category><category>框架</category><category>Agent</category><category>LLM</category><category>MCP</category><category>Skill</category><category>github-weekly</category></item><item><title>上下文工程专题</title><link>https://mingtian-ten.vercel.app/c/agent-interview/context-engineering/</link><guid isPermaLink="true">https://mingtian-ten.vercel.app/c/agent-interview/context-engineering/</guid><description>上下文工程（Context Engineering）面试题整理。从&quot;什么该进上下文&quot;问到压缩策略、位置排布、成本与失败模式，附一次完整的上下文预算分配实例。</description><pubDate>Tue, 29 Sep 2026 08:01:53 GMT</pubDate><category>面试</category><category>上下文工程</category><category>Agent</category><category>agent-interview</category></item><item><title>RAG 专题</title><link>https://mingtian-ten.vercel.app/c/agent-interview/rag/</link><guid isPermaLink="true">https://mingtian-ten.vercel.app/c/agent-interview/rag/</guid><description>RAG 高频面试题整理。从&quot;为什么不是微调&quot;一路问到混合检索权重、重排位置、以及召回率上不去的排查路径。</description><pubDate>Tue, 29 Sep 2026 07:13:00 GMT</pubDate><category>面试</category><category>RAG</category><category>记忆</category><category>agent-interview</category></item><item><title>Function Calling 与工具调用专题</title><link>https://mingtian-ten.vercel.app/c/agent-interview/tool-calling/</link><guid isPermaLink="true">https://mingtian-ten.vercel.app/c/agent-interview/tool-calling/</guid><description>工具调用高频面试题。重点在参数校验、错误回灌、并发与幂等——这些才是把 demo 变成可用系统的分水岭。</description><pubDate>Tue, 29 Sep 2026 06:56:18 GMT</pubDate><category>面试</category><category>Agent</category><category>工具</category><category>agent-interview</category></item><item><title>GitHub 一周热点（第 39 周）</title><link>https://mingtian-ten.vercel.app/c/github-weekly/2026-09-29-week39/</link><guid isPermaLink="true">https://mingtian-ten.vercel.app/c/github-weekly/2026-09-29-week39/</guid><description>本周挑出 5 个项目：两个 Agent 编排框架、一个本地优先的数据工具、一个把 LLM 接进终端的思路，以及一个值得单独读设计的检索库。</description><pubDate>Tue, 29 Sep 2026 06:56:18 GMT</pubDate><category>开源</category><category>工具</category><category>Agent</category><category>github-weekly</category></item></channel></rss>