<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: moin shaikh</title>
    <description>The latest articles on DEV Community by moin shaikh (@moin_shaikh).</description>
    <link>https://dev.to/moin_shaikh</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4245185%2F8345e802-355e-4518-9a57-6cb78859f9d7.jpeg</url>
      <title>DEV Community: moin shaikh</title>
      <link>https://dev.to/moin_shaikh</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/moin_shaikh"/>
    <language>en</language>
    <item>
      <title>I Built Agent Black Box to Investigate What AI Agents Actually Do</title>
      <dc:creator>moin shaikh</dc:creator>
      <pubDate>Sat, 14 Nov 2026 23:19:46 +0000</pubDate>
      <link>https://dev.to/moin_shaikh/i-built-agent-black-box-to-investigate-what-ai-agents-actually-do-2foi</link>
      <guid>https://dev.to/moin_shaikh/i-built-agent-black-box-to-investigate-what-ai-agents-actually-do-2foi</guid>
      <description>&lt;p&gt;I Built Agent Black Box to Investigate What AI Agents Actually Do&lt;/p&gt;

&lt;p&gt;AI agents are becoming increasingly capable. They can call APIs, interact with MCP tools, access resources, and execute multi-step workflows.&lt;/p&gt;

&lt;p&gt;But there’s a security question I’ve been exploring:&lt;/p&gt;

&lt;p&gt;When an AI agent performs an unexpected action, how do we reconstruct what actually happened?&lt;/p&gt;

&lt;p&gt;Traditional application logs can show individual events, but understanding the relationships between agent decisions, tool calls, permissions, and downstream actions is much harder.&lt;/p&gt;

&lt;p&gt;That’s why I’ve been building Agent Black Box (ABB).&lt;/p&gt;

&lt;p&gt;What is Agent Black Box?&lt;/p&gt;

&lt;p&gt;ABB is an observation-only AI agent security forensics platform.&lt;/p&gt;

&lt;p&gt;It helps investigators explore:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Causal security graphs: Visualize relationships between agent actions, tools, identities, and resources.&lt;/li&gt;
&lt;li&gt;MCP tool tracing: Investigate which tools an agent invoked.&lt;/li&gt;
&lt;li&gt;Authority analysis: Examine permission and authority mismatches.&lt;/li&gt;
&lt;li&gt;Incident investigations: Connect detections, timelines, and supporting evidence.&lt;/li&gt;
&lt;li&gt;Multi-agent workflows: Explore behavior across collaborating agents.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The First Hosted Beta Is Live&lt;/p&gt;

&lt;p&gt;After local development and testing, ABB Closed Beta V1 is now running on a VPS with Cloudflare Access authentication.&lt;/p&gt;

&lt;p&gt;The current beta uses synthetic security scenarios to demonstrate investigation workflows.&lt;/p&gt;

&lt;p&gt;It’s important to be clear about the current limitations: ABB is observation-only, single-tenant, and not yet a production-ready multi-tenant security platform. It does not block suspicious actions in real time.&lt;/p&gt;

&lt;p&gt;Looking for Developer Feedback&lt;/p&gt;

&lt;p&gt;I’m particularly interested in hearing from people building AI agents, MCP integrations, and security tooling.&lt;/p&gt;

&lt;p&gt;What evidence would you need to investigate a suspicious AI agent action?&lt;/p&gt;

&lt;p&gt;Website: &lt;a href="https://abbxai.com" rel="noopener noreferrer"&gt;https://abbxai.com&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Status: Closed Beta V1, invite-only access.&lt;/p&gt;

</description>
      <category>security</category>
      <category>ai</category>
      <category>devtools</category>
      <category>mcp</category>
    </item>
  </channel>
</rss>
