A broken access control vulnerability in LibreChat allows authenticated attackers to manipulate agent file contexts, posing a risk of unauthorized system behavior modification.

This vulnerability involves missing authorization controls (CWE-862) regarding file uploads for agent file search and context. An authenticated attacker can exploit this by targeting a known agent ID to upload malicious files, thereby altering the agent's operational logic despite lacking the required administrative permissions.

Successful exploitation allows an attacker to manipulate the responses and behaviors of AI agents within the LibreChat platform. Given the CVSS score of 7.1, this represents a high-severity risk that could lead to the subversion of internal workflows, the injection of malicious instructions into automated processes, and potential reputational damage if the AI agent is used for client-facing communications.

Immediate Action: Upgrade your LibreChat deployment to version 0.8.2-rc2 or later, which includes the necessary authorization checks to prevent unauthorized file uploads.

Proactive Monitoring: Review audit logs for unusual file upload activity or modifications to agent configurations, specifically monitoring for actions performed by users who do not typically manage those specific agents.

Compensating Controls: If immediate patching is not feasible, restrict access to agent management features at the application level and implement strict egress filtering to prevent agents from interacting with unauthorized external file sources.

Public Exploit Available: Unknown

The ability for a standard user to manipulate AI agent behavior constitutes a significant integrity risk. Security teams should prioritize the update to version 0.8.2-rc2 immediately to enforce proper access control and prevent unauthorized modification of agent file contexts.