<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Tag-Injection on i break software - My work with different software, bug hunting and interesting tidbits.</title>
    <link>/tags/tag-injection/</link>
    <description>Recent content in Tag-Injection on i break software - My work with different software, bug hunting and interesting tidbits.</description>
    <generator>Hugo</generator>
    <language>en-us</language>
    <lastBuildDate>Tue, 20 Oct 2026 14:29:37 +0530</lastBuildDate>
    <atom:link href="/tags/tag-injection/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Closing the Tag, 0wning the Agent: Tag Boundary Injection in Bedrock Agents</title>
      <link>/2026/10/closing-the-tag-0wning-the-agent-c0c0n-2026/</link>
      <pubDate>Sat, 10 Oct 2026 00:00:00 +0000</pubDate>
      <guid>/2026/10/closing-the-tag-0wning-the-agent-c0c0n-2026/</guid>
      <description>&lt;p&gt;Slides and a tool release from my talk at c0c0n 2026 on the possibilities of tag boundary injection in AWS Bedrock Agent orchestration. The research attempted to answer whether an attacker who controls input can close, duplicate or forge the &lt;code&gt;&amp;lt;thinking&amp;gt;&lt;/code&gt;/&lt;code&gt;&amp;lt;answer&amp;gt;&lt;/code&gt; tags that Bedrock uses to drive an agent. The findings, and how model dependent the approach can be, are covered inside the post. Also releasing &lt;strong&gt;AWS Bedrock TagHammer&lt;/strong&gt;, the harness/fuzzer I built to test all of this.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
