Superagi EUVD-2026-23719

Exploitation requires valid authentication to SuperAGI (PR:L in CVSS vector) - the attacker must possess a legitimate user account or obtain one through credential compromise. No special network conditions, non-default configurations, or user interaction are required beyond initial authentication. The API Key Management endpoints must be accessible over the network (AV:N). The vulnerability affects all authenticated users equally, including low-privilege accounts, with no role-based restrictions preventing lateral key manipulation. However, exploitation is limited to modification of API keys themselves (not full system compromise) - the attacker cannot escalate to administrative access or read credentials without further attacks.

CVSS 5.4 (Medium) with network attack vector and low complexity reflects moderate real-world risk. The CVSS vector includes PR:L (requires authentication), AC:L (low attack complexity), and E:P (exploit proof-of-concept exists), indicating this is reliably exploitable by any authenticated user without special conditions. However, the impact is limited to integrity and availability of API keys (I:L/A:L) rather than confidentiality, reducing severity. The absence of KEV listing despite public exploit availability suggests this is not yet part of widespread active exploitation campaigns, though targeted exploitation is probable given the ease of execution. The vendor's non-response to early disclosure indicates no patched version is available, elevating practical risk for organizations running affected versions. This represents a real but not critical priority - attackers need valid authentication first, but any authenticated user can then compromise other users' API credentials.

An authenticated user with valid credentials (e.g., a low-privilege employee or attacker with a compromised account) sends a crafted HTTP request to the delete_api_key or edit_api_key endpoint specifying another user's API key ID. Due to missing authorization checks, the endpoint processes the request and modifies or deletes the target API key, allowing the attacker to disable a colleague's integrations or steal API credentials by replacing them with controlled values. Publicly available exploit code demonstrates this by iterating through API key IDs to identify and hijack credentials.