Ruflo MCP Flaw Lets Unauthenticated Attackers Run Commands and Poison AI Memory

A severe security flaw has been discovered in Ruflo, an open-source agent meta-harness for certain AI models. The vulnerability allows unauthenticated attackers to execute code remotely. It affects all versions of the project prior to version 3.16.3 and has been assigned a CVSS score of 10.0. The issue has been dubbed RufRoot by Noma Security's researchers. This flaw could potentially enable attackers to run commands and manipulate AI memory. The vulnerability is tracked as CVE-2026-59726.