<?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>Prompt-Injection on 兜忆轩 | Cedric's Compiler Notes</title><link>https://douyixuan.github.io/tags/prompt-injection/</link><description>Recent content in Prompt-Injection on 兜忆轩 | Cedric's Compiler Notes</description><image><title>兜忆轩 | Cedric's Compiler Notes</title><url>https://douyixuan.github.io/images/papermod-cover.png</url><link>https://douyixuan.github.io/images/papermod-cover.png</link></image><generator>Hugo -- 0.146.0</generator><language>en-us</language><lastBuildDate>Fri, 11 Sep 2026 08:00:00 +0800</lastBuildDate><atom:link href="https://douyixuan.github.io/tags/prompt-injection/index.xml" rel="self" type="application/rss+xml"/><item><title>AI Native Digest 2026-09-11</title><link>https://douyixuan.github.io/posts/ai-native/2026-09-11/</link><pubDate>Fri, 11 Sep 2026 08:00:00 +0800</pubDate><guid>https://douyixuan.github.io/posts/ai-native/2026-09-11/</guid><description>今天关注 AI-native Agent 的两个结构性可靠性问题：Meta 用 Secure VM、credential isolation 和独立 Sentinel 把个人 Agent 的权限控制移出模型；ExecCritic 则证明 coding agent 的执行反馈只有在测试本身可信时才有价值，并用独立 Test Agent + fail-closed harness 避免 Agent 自己给自己验收。</description></item></channel></rss>