An improper input validation flaw in the IBM Langflow OSS PythonREPL sandbox allows an authenticated user to perform code injection, leading to potential remote code execution.

The vulnerability is a code injection flaw (CWE-94) within the PythonREPL sandbox. It requires the attacker to be authenticated to interact with the vulnerable component, where they can then bypass input validation to execute arbitrary code.

With a CVSS score of 9.9, this vulnerability represents a severe risk to the integrity and security of the application environment. An attacker who has gained authenticated access can move beyond the intended sandbox restrictions to run arbitrary commands, potentially leading to full control over the application's runtime environment and sensitive data access.

Immediate Action: Upgrade IBM Langflow OSS to version 1.10.2 or later to address the sandbox validation flaw.

Proactive Monitoring: Review application logs for suspicious Python code execution patterns or unexpected shell commands originating from the Langflow interface.

Compensating Controls: Implement strict access control lists to limit which users can interact with the PythonREPL feature, and ensure the application is running with the principle of least privilege.

Public Exploit Available: Unknown

Organizations using Langflow OSS must update to version 1.10.2 immediately to remediate the sandbox bypass. Limiting access to the application and monitoring for unusual activity is recommended until the patch is successfully deployed.