Use GitHub's private vulnerability reporting on this repository (Security → Report a vulnerability). Please do not open a public issue for anything exploitable.

Two different kinds of finding land here, and both are welcome:

• Code vulnerabilities — in the gateway (enterprise/gateway/hub.py), the worker agent, or the scripts. The gateway is the deliberate attack surface: it accepts workflow JSON from users and serves files from a shared volume, so path handling and input validation bugs there matter most. Under AUTH_MODE=oauth it scopes /outputs to the caller's own workspace and /api/showback to the caller's own row; a way past either scope, or a path ComfyUI reports that reaches a browser unconfined, is in scope here.
• Deployment-pattern weaknesses — a manifest or script in this repo that configures something insecurely by default. The security model this repo promises is documented in docs/04-exposing.md and docs/06-enterprise-architecture.md (short version: raw ComfyUI is never network-reachable; the GPU pods bind loopback and have no Service or Route; the namespace is default-deny and a GPU pod can reach only Redis and DNS, as the comfy-worker ACL user and nothing broader). Anything that quietly breaks that promise is a vulnerability, not a nitpick.

ComfyUI itself and its custom-node ecosystem are upstream projects — report their vulnerabilities upstream. This repo's mitigation for that entire class is architectural: keep ComfyUI unreachable and ENABLE_MANAGER=false.