<?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>Forem: KnowTechie</title>
    <description>The latest articles on Forem by KnowTechie (@knowtechie).</description>
    <link>https://forem.com/knowtechie</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.amazonaws.com%2Fuploads%2Forganization%2Fprofile_image%2F8401%2F80f29813-a628-488c-9181-4713b1278117.png</url>
      <title>Forem: KnowTechie</title>
      <link>https://forem.com/knowtechie</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://forem.com/feed/knowtechie"/>
    <language>en</language>
    <item>
      <title>GitHub’s New MCP Can Spill Your Secrets—No Hacking Required</title>
      <dc:creator>Kevin Raposo</dc:creator>
      <pubDate>Tue, 27 May 2025 15:46:59 +0000</pubDate>
      <link>https://forem.com/knowtechie/githubs-new-mcp-can-spill-your-secrets-no-hacking-required-51n2</link>
      <guid>https://forem.com/knowtechie/githubs-new-mcp-can-spill-your-secrets-no-hacking-required-51n2</guid>
      <description>&lt;p&gt;GitHub’s Model Context Protocol (MCP) just landed in hot water, thanks to a &lt;a href="https://simonwillison.net/2025/May/26/github-mcp-exploited/" rel="noopener noreferrer"&gt;newly discovered vulnerability&lt;/a&gt; that lets attackers trick AI agents into leaking private repository information. &lt;/p&gt;

&lt;p&gt;Security researchers Marco Milanta and Luca Beurer-Kellner stumbled on an exploit where an attacker can file a sneaky issue in a public repo. &lt;/p&gt;

&lt;p&gt;If a user asks an LLM agent connected to MCP to “check the issues,” the agent follows the attacker’s instructions—like digging into all the user’s private repos—and then exposes that info in a public pull request. &lt;/p&gt;

&lt;p&gt;No malware, no brute force, just a well-crafted prompt and a bit of bad architecture.&lt;/p&gt;

&lt;p&gt;Here’s the kicker: this isn’t a bug in the code—it’s a design flaw&lt;/p&gt;

&lt;h3&gt;
  
  
  The holy trinity for prompt injection attacks
&lt;/h3&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fxa6cf1ua3kdewybaaxd6.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fxa6cf1ua3kdewybaaxd6.png" alt="Image image showing prompt to hack Github MCP" width="800" height="483"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;According to &lt;a href="https://devclass.com/2025/05/27/researchers-warn-of-prompt-injection-vulnerability-in-github-mcp-with-no-obvious-fix/" rel="noopener noreferrer"&gt;DevClass&lt;/a&gt;, the MCP server gives LLMs access to private data, lets them process attacker-controlled prompts, and allows them to exfiltrate information, all at once. &lt;/p&gt;

&lt;p&gt;&lt;a href="https://invariantlabs.ai/blog/mcp-github-vulnerability" rel="noopener noreferrer"&gt;Security folks are already warning&lt;/a&gt; that there’s no obvious fix in sight. The only advice? If you’re using MCP, treat it like a loaded gun around anything private. &lt;/p&gt;

&lt;p&gt;The attack doesn’t require elite skills—just a clever issue and a bit of trust in the wrong place (Invariant Labs).&lt;/p&gt;

&lt;p&gt;As of now, GitLab hasn’t released an official statement or any mitigation. So if you’re experimenting with MCP, keep your secrets close and your AI agents on a tight leash.&lt;/p&gt;

</description>
      <category>github</category>
      <category>ai</category>
      <category>llm</category>
      <category>mcp</category>
    </item>
    <item>
      <title>Powerful AI tool Actionize connects GPT to everything</title>
      <dc:creator>Kevin Raposo</dc:creator>
      <pubDate>Fri, 05 Apr 2024 16:48:53 +0000</pubDate>
      <link>https://forem.com/knowtechie/powerful-ai-tool-actionize-connects-gpt-to-everything-2gg4</link>
      <guid>https://forem.com/knowtechie/powerful-ai-tool-actionize-connects-gpt-to-everything-2gg4</guid>
      <description>&lt;p&gt;From the creator of notable tools like Chatwith, Spell.so and Gobblebot, indie hacker &lt;a class="mentioned-user" href="https://dev.to/rafalzawadzki"&gt;@rafalzawadzki&lt;/a&gt; is at it once again. Today, we spotlight his latest unveiling — Actionize, a powerful AI tool set to change how your GPT communicates… with everything.&lt;/p&gt;

&lt;p&gt;Imagine if your GPT could shoot off emails in Gmail, send newsletters with Mailchimp, and dig out info from Google Sheets. It’s not just talk – with Actionize, it’s a reality and can connect your GPT to pretty much anything.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://knowtechie.com/actionize-gpt-ai-tool/"&gt;Click here&lt;/a&gt; to read about how it works, how much it costs, and more. &lt;/p&gt;

</description>
      <category>chatgpt</category>
      <category>ai</category>
      <category>productivity</category>
      <category>news</category>
    </item>
  </channel>
</rss>
