The vulnerability in the execute_code function of eosphoros‑ai DB‑GPT’s CodeAction component allows an attacker to inject arbitrary code. By passing specially crafted input to the vulnerable code_utils.py routine, command and code injection (CWE‑74 and CWE‑94) can occur, giving the attacker the ability to run arbitrary commands on the host. This weakness can compromise confidentiality, integrity, and availability of the affected system, and it is activated through a remote input vector.

The flaw affects eosphoros‑ai DB‑GPT versions up to and including 0.8.2. The injected code is executed within the context of the application’s runtime, so only systems running the vulnerable release and using the CodeAction execute_code feature are impacted.

The CVSS score of 6.9 classifies the risk as moderate. The exploit is publicly available and has been released to the public, indicating that an attacker has already demonstrated the technique. Although the EPSS score is not available, the existence of a released exploit and the ability to trigger it remotely increases the likelihood of real‑world attacks. The vulnerability is not listed in the CISA KEV catalog, but the lack of a formal vendor response suggests a potential window for mitigation. The attack vector is remote and can be invoked by sending malicious input to the execute_code endpoint, which then executes the payload without adequate input validation.