Executive brief
PraisonAI Platform, a system for managing multi-agent AI teams, contains a security flaw in how it handles task dependencies. An authorized user in one workspace can view, delete, or create links between tasks (issues) in other workspaces where they have no permissions. This could lead to unauthorized data access, disruption of project workflows, and the ability to manipulate the relationship between different AI-driven tasks across the entire platform.
Technical details
An Insecure Direct Object Reference (IDOR) exists in the dependency endpoints of the PraisonAI Platform. The application validates that a user belongs to the `workspace_id` provided in the URL but fails to verify if the `issue_id` or `dep_id` actually belong to that workspace. Specifically, the `DependencyService` methods (`create`, `list_for_issue`, and `delete`) query the database using raw IDs without workspace-scoping predicates. Most critically, the `create_dependency` endpoint accepts a `depends_on_issue_id` in the request body that is never validated, allowing an attacker to create "blocks" or "related" links between any two issues in the database. This allows for cross-workspace unauthorized read, delete, and creation of issue dependencies. The issue is fixed in version 0.1.4.
Affected products
- MervinPraison PraisonAI Platform < 0.1.4
Timeline
- 2026-05-19: patched: Fix merged in commit ef79b7a
- 2026-07-21: disclosed: CVE-2026-47406 published