ComfyUI-Easy-Use flaw allows unauthenticated remote code execution

A path traversal bug in the ComfyUI-Easy-Use extension lets attackers overwrite a custom node and run arbitrary code when ComfyUI restarts.

On November 12, 2026, GitHub Advisory Database published the GHSA-25xv-w465-qr9f advisory describing a path traversal flaw in ComfyUI-Easy-Use, a custom-node extension for the open source ComfyUI image and video generation tool. The flaw lets an unauthenticated, network-reachable attacker achieve remote code execution.

What the advisory describes

Tracked as CVE-2026-123790 and GHSA-25xv-w465-qr9f, published November 12, 2026
The flaw sits in the save_preview() handler for POST requests to /easyuse/save/{name}
Insufficient filename validation lets an attacker point the write destination at the custom_nodes directory
An unauthenticated, network-reachable attacker can overwrite an installed custom node's Python entry point with an uploaded file
Arbitrary code execution happens with the ComfyUI process's privileges once the modified node loads on restart
The affected version is confirmed as 1.3.6; the advisory does not specify a patched version

What it means for people who build with AI

ComfyUI is one of the most widely deployed open source AI image and video generation pipelines, and this flaw affects a popular extension within that ecosystem. Because the attacker needs no credentials and only network access to the ComfyUI instance, self-hosted creative AI setups that expose ComfyUI-Easy-Use are at risk of having their custom nodes silently replaced and executed with the privileges of the ComfyUI process.

Availability of a fix

The advisory confirms version 1.3.6 of ComfyUI-Easy-Use is affected but does not specify which version carries a fix. Anyone running this extension should check their installed version against the advisory and watch for an updated release before continuing to expose ComfyUI-Easy-Use over a network.

Sources

GHSA-25xv-w465-qr9f (GitHub Advisory Database, November 12, 2026, official)

Written with AI assistance, strictly from the sources listed above. Before it went live, automated checks verified its dates, sources and every language edition.