A flaw in the function that extracts command data from arguments in the backend server of the gpt-researcher application allows a malicious actor to inject arbitrary code. The injection occurs when attacker-controlled input is passed to the server endpoint, resulting in execution of unintended commands. This could compromise the confidentiality, integrity, and availability of the host system.

The vulnerability affects versions of assafelovic's gpt-researcher up to and including 3.4.3. No later releases are confirmed to be fixed in the provided data.

The CVSS base score of 6.9 indicates a significant risk level. Though no EPSS score is available and the flaw is not listed in CISA's KEV catalog, the public disclosure and the ability to trigger the issue from a remote source suggest that an attacker could exploit the flaw without local privileges. The attack likely requires sending crafted payloads to the websocket endpoint’s extract_command_data function, which the application currently sanitizes inadequately. Without an official patch, the risk remains present until mitigation steps are applied.