Executive brief
Flowise, an open-source tool for building LLM applications, contains a flaw in how it handles deletion requests for chat and agent workflows. An authorized user with permission to delete only one type of workflow can exploit this bug to delete the other type. This can lead to the unauthorized removal of critical configurations and service disruptions within a shared workspace.
Technical details
The vulnerability exists in the `DELETE /api/v1/chatflows/:id` endpoint, which uses `checkAnyPermission('chatflows:delete,agentflows:delete')` for authorization. While the endpoint correctly checks for the presence of either permission, the underlying logic fails to verify that the target resource's type matches the specific permission held by the requester. An attacker with low-privileged access (possessing only one of the two delete permissions) can provide the ID of a resource belonging to the other category to successfully delete it. This results in a breakdown of Role-Based Access Control (RBAC) separation between Chatflows and Agentflows. The issue is addressed in version 3.1.3.
Affected products
- FlowiseAI flowise <= 3.1.2
Timeline
- 2026-07-29: disclosed: Initial report to vendor
- 2026-08-04: advisory: GitHub Advisory published
- 2026-08-04: patched: Version 3.1.3 released