ComfyGuard is a security tool, so its own security matters. This policy covers vulnerabilities in ComfyGuard itself.

Please report privately, not in a public issue.

• Preferred: use GitHub's private vulnerability reporting on this repository (the Security tab, then Report a vulnerability).
• Alternatively, contact the maintainer (@joosthel) privately.

Include what you found, how to reproduce it, and the impact. We aim to acknowledge a report within a few days and to keep you updated as we work on a fix. Please give us reasonable time to address the issue before any public disclosure.

In scope is anything in this repository: the scanner and its logic, the report handling, the ruleset, and the agent skills. Because ComfyGuard is strictly read-only by design, the issues we care most about are:

• A path where the tool could change a scanned instance at all (it must only ever write its own report artifacts).
• A path where scanning a hostile target (a crafted node, model, or workflow) could execute code, read data outside the scan, or exfiltrate anything from the host running ComfyGuard.
• Secret values leaking into a report (it is designed to record names and locations only, never secret values).

• Vulnerabilities in ComfyUI itself or in ComfyUI-Manager. Report those to their maintainers through the ComfyUI security process. ComfyGuard reports on such issues but does not fix them.
• Vulnerabilities in third-party custom nodes. Report those to the node's author.
• A clean grade or a missed finding. ComfyGuard does not claim to prove an instance is safe; a missed detection is a bug or a feature request, not a vulnerability in ComfyGuard.

ComfyGuard is a defensive self-assessment tool. Use it only against ComfyUI deployments you operate or are authorized to test. The active network probe is opt-in, defaults to localhost, and requires you to assert authorization for any non-loopback target.