<?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: Bindfort</title>
    <description>The latest articles on DEV Community by Bindfort (@oleg_gr_734317a4bae97cee4).</description>
    <link>https://dev.to/oleg_gr_734317a4bae97cee4</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%2Fuser%2Fprofile_image%2F3929315%2Fe412e153-73f3-4ac5-a17e-6a8a5d2eb978.png</url>
      <title>DEV Community: Bindfort</title>
      <link>https://dev.to/oleg_gr_734317a4bae97cee4</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/oleg_gr_734317a4bae97cee4"/>
    <language>en</language>
    <item>
      <title>Your MCP dependency scan can pass and still miss HIGH vulnerabilities</title>
      <dc:creator>Bindfort</dc:creator>
      <pubDate>Wed, 13 May 2026 13:36:32 +0000</pubDate>
      <link>https://dev.to/oleg_gr_734317a4bae97cee4/your-mcp-dependency-scan-can-pass-and-still-miss-high-vulnerabilities-54m8</link>
      <guid>https://dev.to/oleg_gr_734317a4bae97cee4/your-mcp-dependency-scan-can-pass-and-still-miss-high-vulnerabilities-54m8</guid>
      <description>&lt;p&gt;Quick story, then the practical part.&lt;/p&gt;

&lt;p&gt;We scanned five official MCP reference servers from the &lt;code&gt;@modelcontextprotocol&lt;/code&gt; npm namespace. Standard tooling against the package manifest:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;0 findings
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Then we re-ran the same check against the installed dependency tree:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;10 HIGH findings
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Same five servers. Same advisory database. The difference was that the second scan walked into a package the first one never had reason to query: &lt;code&gt;@modelcontextprotocol/sdk@1.0.1&lt;/code&gt;.&lt;/p&gt;

&lt;p&gt;The advisories were public. The fixes were already shipped. The scan just didn't reach that far down.&lt;/p&gt;

&lt;h2&gt;
  
  
  What we scanned
&lt;/h2&gt;

&lt;p&gt;The five official reference servers, on April 26, 2026:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;&lt;code&gt;@modelcontextprotocol/server-filesystem&lt;/code&gt;&lt;/li&gt;
&lt;li&gt;&lt;code&gt;@modelcontextprotocol/server-github&lt;/code&gt;&lt;/li&gt;
&lt;li&gt;&lt;code&gt;@modelcontextprotocol/server-everything&lt;/code&gt;&lt;/li&gt;
&lt;li&gt;&lt;code&gt;@modelcontextprotocol/server-memory&lt;/code&gt;&lt;/li&gt;
&lt;li&gt;&lt;code&gt;@modelcontextprotocol/server-sequential-thinking&lt;/code&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The scan walked the full installed tree using &lt;code&gt;npm ls --json --all&lt;/code&gt;, then sent every resolved package/version pair through OSV.dev.&lt;/p&gt;

&lt;h2&gt;
  
  
  What turned up
&lt;/h2&gt;

&lt;p&gt;Every server resolved the same SDK version:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;@modelcontextprotocol/sdk@1.0.1
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;That version sits behind two public HIGH advisories:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Advisory&lt;/th&gt;
&lt;th&gt;Issue&lt;/th&gt;
&lt;th&gt;Fixed in&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;GHSA-8r9q-7v3j-jr4g&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;ReDoS in the SDK parser&lt;/td&gt;
&lt;td&gt;&lt;code&gt;@modelcontextprotocol/sdk@1.25.2&lt;/code&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;GHSA-w48q-cv73-mx4w&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;DNS rebinding from missing default &lt;code&gt;Host&lt;/code&gt; validation&lt;/td&gt;
&lt;td&gt;&lt;code&gt;@modelcontextprotocol/sdk@1.24.0&lt;/code&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Two advisories × five servers = ten HIGH findings.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why these matter (more than usual)
&lt;/h2&gt;

&lt;p&gt;The ReDoS one is the interesting one. It's triggerable through an MCP tool &lt;em&gt;response&lt;/em&gt;, not through a normal inbound HTTP request. A compromised or malicious upstream server can return crafted content and pin the calling agent's Node.js event loop. That's the opposite direction from where most application security tooling looks.&lt;/p&gt;

&lt;p&gt;The DNS rebinding one is more familiar but still serious in this context. A user has an MCP server running locally. They open a browser tab. After DNS rebinding, attacker-controlled JavaScript can reach the local MCP server and call exposed tools. Filesystem, shell, repo, database — whatever the server hands out.&lt;/p&gt;

&lt;p&gt;No malware. No privilege escalation. A browser tab and a service that didn't validate &lt;code&gt;Host&lt;/code&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why the shallow scan returned clean
&lt;/h2&gt;

&lt;p&gt;The &lt;code&gt;@modelcontextprotocol/server-*&lt;/code&gt; packages don't have GHSA entries on themselves. Ask a top-level scanner about them and you'll get back exactly what you asked for:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;@modelcontextprotocol/server-filesystem  →  0 findings
@modelcontextprotocol/server-github      →  0 findings
@modelcontextprotocol/server-everything  →  0 findings
@modelcontextprotocol/server-memory      →  0 findings
@modelcontextprotocol/server-sequential  →  0 findings
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;What's actually on disk looks more like this:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;@modelcontextprotocol/server-filesystem
  └─ @modelcontextprotocol/sdk@1.0.1
       ├─ GHSA-8r9q-7v3j-jr4g
       └─ GHSA-w48q-cv73-mx4w
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Same database, same packages, different scan depth.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;Shallow package scan:  5 × 0  =  0 findings
Recursive tree scan:   5 × 2  = 10 HIGH
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h2&gt;
  
  
  The check you can run right now
&lt;/h2&gt;

&lt;p&gt;From inside the project (or installed server directory):&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;npm &lt;span class="nb"&gt;ls&lt;/span&gt; @modelcontextprotocol/sdk &lt;span class="nt"&gt;--all&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Installed SDK version&lt;/th&gt;
&lt;th&gt;Status&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;&amp;lt;1.24.0&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;Both advisories present&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;
&lt;code&gt;1.24.0&lt;/code&gt; through &lt;code&gt;1.25.1&lt;/code&gt;
&lt;/td&gt;
&lt;td&gt;DNS rebinding patched; ReDoS still present&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;&amp;gt;=1.25.2&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;Both advisories patched&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;If the lockfile still pins &lt;code&gt;1.0.1&lt;/code&gt;, just bumping &lt;code&gt;package.json&lt;/code&gt; won't help. Update the lockfile, reinstall, and confirm with &lt;code&gt;npm ls&lt;/code&gt; that the resolved version actually changed.&lt;/p&gt;

&lt;h2&gt;
  
  
  What a better scanner has to do
&lt;/h2&gt;

&lt;p&gt;For npm-based MCP servers, the useful target is the installed tree:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;npm &lt;span class="nb"&gt;ls&lt;/span&gt; &lt;span class="nt"&gt;--json&lt;/span&gt; &lt;span class="nt"&gt;--all&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Then walk it recursively and query every node:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight go"&gt;&lt;code&gt;&lt;span class="k"&gt;func&lt;/span&gt; &lt;span class="n"&gt;walkNPMTree&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;tree&lt;/span&gt; &lt;span class="k"&gt;map&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="kt"&gt;string&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="n"&gt;npmTreeNode&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;out&lt;/span&gt; &lt;span class="o"&gt;*&lt;/span&gt;&lt;span class="p"&gt;[]&lt;/span&gt;&lt;span class="n"&gt;DependencyInput&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
    &lt;span class="k"&gt;for&lt;/span&gt; &lt;span class="n"&gt;name&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;node&lt;/span&gt; &lt;span class="o"&gt;:=&lt;/span&gt; &lt;span class="k"&gt;range&lt;/span&gt; &lt;span class="n"&gt;tree&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
        &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="n"&gt;name&lt;/span&gt; &lt;span class="o"&gt;!=&lt;/span&gt; &lt;span class="s"&gt;""&lt;/span&gt; &lt;span class="o"&gt;&amp;amp;&amp;amp;&lt;/span&gt; &lt;span class="n"&gt;node&lt;/span&gt;&lt;span class="o"&gt;.&lt;/span&gt;&lt;span class="n"&gt;Version&lt;/span&gt; &lt;span class="o"&gt;!=&lt;/span&gt; &lt;span class="s"&gt;""&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
            &lt;span class="o"&gt;*&lt;/span&gt;&lt;span class="n"&gt;out&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nb"&gt;append&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;*&lt;/span&gt;&lt;span class="n"&gt;out&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;DependencyInput&lt;/span&gt;&lt;span class="p"&gt;{&lt;/span&gt;
                &lt;span class="n"&gt;Name&lt;/span&gt;&lt;span class="o"&gt;:&lt;/span&gt;      &lt;span class="n"&gt;name&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
                &lt;span class="n"&gt;Version&lt;/span&gt;&lt;span class="o"&gt;:&lt;/span&gt;   &lt;span class="n"&gt;node&lt;/span&gt;&lt;span class="o"&gt;.&lt;/span&gt;&lt;span class="n"&gt;Version&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
                &lt;span class="n"&gt;Ecosystem&lt;/span&gt;&lt;span class="o"&gt;:&lt;/span&gt; &lt;span class="s"&gt;"npm"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
            &lt;span class="p"&gt;})&lt;/span&gt;
        &lt;span class="p"&gt;}&lt;/span&gt;
        &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="nb"&gt;len&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;node&lt;/span&gt;&lt;span class="o"&gt;.&lt;/span&gt;&lt;span class="n"&gt;Dependencies&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;&amp;gt;&lt;/span&gt; &lt;span class="m"&gt;0&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
            &lt;span class="n"&gt;walkNPMTree&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;node&lt;/span&gt;&lt;span class="o"&gt;.&lt;/span&gt;&lt;span class="n"&gt;Dependencies&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;out&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
        &lt;span class="p"&gt;}&lt;/span&gt;
    &lt;span class="p"&gt;}&lt;/span&gt;
&lt;span class="p"&gt;}&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;For a typical MCP server install that yields around 179 dependency records. Send them through OSV's batch endpoint and the SDK shows up as one of the records that returns hits.&lt;/p&gt;

&lt;p&gt;The &lt;code&gt;--all&lt;/code&gt; flag matters: without it, npm collapses deduped packages and you can lose parts of the resolved tree.&lt;/p&gt;

&lt;h2&gt;
  
  
  Closing note
&lt;/h2&gt;

&lt;p&gt;This isn't a zero-day. The advisories are published, the fixes exist, and we're not the first to notice either one.&lt;/p&gt;

&lt;p&gt;The point is more boring and probably more useful: known-vulnerable code can sit one level below the package your scanner checked, and "scan passed" doesn't always mean "tree is clean." For MCP, that's worth getting right before pointing an agent at production.&lt;/p&gt;

&lt;p&gt;Full write-up and contact: &lt;a href="https://bindfort.io/blog/mcp-transitive-cve-scan-2026" rel="noopener noreferrer"&gt;https://bindfort.io/blog/mcp-transitive-cve-scan-2026&lt;/a&gt;&lt;/p&gt;

</description>
      <category>security</category>
      <category>mcp</category>
      <category>npm</category>
      <category>appsec</category>
    </item>
  </channel>
</rss>
