mcp-ffmpeg executes system shell commands to manipulate media inputs, but it wrongly concatenates user-supplied videoPath or outputDir values into a string passed to child_process.exec. An attacker can embed characters such as $(), backticks, or a closing double quote to terminate the intended command and inject arbitrary shell instructions. This flaw allows the attacker to run commands as the MCP server user, granting read/write access to the server filesystem and the ability to execute code with the privileges the server process holds.

Bitscorp's MCP-FFmpeg library, which powers media conversion in Minimum Control Platform deployments, is affected in versions up to and including 1.0.5. Any installation that uses the resize‑video, extract‑audio, or get‑video‑info commands with this version should be considered at risk.

The CVSS score of 7.3 indicates high severity. No EPSS score is available, and the vulnerability is not listed in the CISA KEV catalogue. The attack vector is likely remote, as any actor who can influence the videoPath or outputDir parameters—through uploads, API requests, or file‑system manipulation—can exploit the flaw. Successful exploitation yields remote code execution with the same privileges as the MCP server process, potentially compromising the entire host. Immediate patching or mitigations are strongly advised.