The vulnerability allows an attacker to craft a prompt that tricks the HAI Build Code Generator into treating a malicious command as safe. By wrapping the command in a template that the model interprets as harmless, the system bypasses the user approval step and executes the command automatically. This results in arbitrary command execution, compromising confidentiality, integrity, and availability of the executing environment.

The effect is confined to systems that host the HAI Build Code Generator, an automated terminal command executor used for code generation. No specific vendor, product name, or version information is listed, so all installations of this component are potentially affected until a patch is applied or controls are implemented.

No EPSS score or KEV status is available, leaving the specific exploitation likelihood unclear. Nonetheless, the flaw permits remote code execution with no user interaction required once the prompt injection succeeds, indicating a high severity risk. Since the exploit depends mainly on injecting malicious output into the model’s prompt, it can be performed by any actor who can influence the input to the generator, making the risk readily exploitable under the conditions described.