An unauthenticated remote injection vulnerability in PromtEngineer localGPT allows remote attackers to compromise the LLM Prompt Handler component.

The flaw exists in the _route_using_overviews function within backend/server.py, which fails to properly neutralize input, resulting in an injection vulnerability that can be triggered by an unauthenticated remote attacker.

The vulnerability carries a CVSS score of 7.3, indicating a high severity risk. Successful exploitation could allow unauthorized remote manipulation of the LLM Prompt Handler, potentially leading to data integrity issues or unauthorized system interaction. Given that the software lacks a formal versioning system and the vendor has been unresponsive, the risk of unpatched exposure is significant.

Immediate Action: Since no official patch is available, organizations should restrict network access to the localGPT instance, ensuring it is not exposed to untrusted networks or the public internet.

Proactive Monitoring: Monitor server logs for unusual patterns or unexpected input strings directed at the backend/server.py endpoint to identify potential exploitation attempts.

Compensating Controls: Deploy a Web Application Firewall (WAF) with strict input validation rules to filter malicious payloads directed at the vulnerable component.

Public Exploit Available: Yes, a published proof-of-concept exists and is attributed to the security researcher's disclosure at https://github.com/August829/CVEP/issues/9.

Due to the lack of vendor-supplied patches and the presence of a public proof-of-concept, this vulnerability poses a credible risk to any environment hosting localGPT. Security teams must isolate the application immediately and implement strict network access controls to prevent unauthorized remote exploitation. Continued monitoring for anomalous traffic is essential until a permanent fix or transition to a more secure alternative can be implemented.