A weakness in InternScience InternAgent 1.5.0 arises from how the train function in experiment.py handles the argument base_model_name_or_path. The function uses this parameter to include and execute code from the specified path. An attacker can supply a malicious value that points to an untrusted source, resulting in the application importing and running arbitrary code. The description states that the attack may be initiated remotely and that a public exploit is available, indicating that the flaw can be leveraged without local access.

The only product affected is InternScience InternAgent version 1.5.0. No other vendors or product versions are listed. The vulnerability manifests when the train functionality processes a crafted base_model_name_or_path value, allowing inclusion of external code. Users deploying 1.5.0 who invoke the train operation are therefore exposed.

The CVSS score of 5.3 classifies this as a moderate severity issue. EPSS is not available, so the current exploitation probability remains unknown. The flaw is not yet listed in the CISA KEV catalog, yet a publicly distributed proof‑of‑concept indicates that exploitation is feasible. Because the attack vector is remote and the exploit does not require elevated privileges, the risk to affected systems is significant. The absence of a known KEV listing does not diminish the urgency, given the availability of an exploit and the potential for remote code execution.