<?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>Agentevals on Thomas Talks AI: Claude Architecture, AI Agent Security, Claude Code &amp; Production Engineering</title>
    <link>https://thomasdevos.com/tags/agentevals/</link>
    <description>Recent content in Agentevals on Thomas Talks AI: Claude Architecture, AI Agent Security, Claude Code &amp; Production Engineering</description>
    <generator>Hugo -- 0.145.0</generator>
    <language>en-us</language>
    <copyright>Thomas De Vos ©</copyright>
    <lastBuildDate>Thu, 23 Jul 2026 10:44:00 +0100</lastBuildDate>
    <atom:link href="https://thomasdevos.com/tags/agentevals/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Test the tool path before you trust the agent&#39;s answer</title>
      <link>https://thomasdevos.com/posts/2026-07-23.10.44.00-test-the-tool-path-before-you-trust-the-agent-answer/</link>
      <pubDate>Thu, 23 Jul 2026 10:44:00 +0100</pubDate>
      <guid>https://thomasdevos.com/posts/2026-07-23.10.44.00-test-the-tool-path-before-you-trust-the-agent-answer/</guid>
      <description>Learn how to test AI agent tool calls, identity, data access, approvals, side effects, and rollback instead of scoring only the final answer.</description>
    </item>
  </channel>
</rss>
