This vulnerability affects FoundationAgents MetaGPT versions 0.8.0 through 0.8.2. It arises from unsafe deserialization in the Message.check_instruct_content function, allowing an attacker with local access to manipulate input arguments to trigger deserialization. The vulnerability does not require user interaction and has low complexity but is limited to local privilege levels. No official remediation or patch has been released as of the publication date. The CVSS 4.0 base score is 4.8, reflecting medium severity with local attack vector and limited impact scope.

Successful exploitation could allow an attacker with local access and low privileges to execute deserialization, potentially leading to unauthorized code execution or data manipulation within the affected application context. However, the attack cannot be performed remotely and requires local access, limiting the overall risk. There are no reports of active exploitation in the wild.

No official patch or remediation has been published by the vendor. Since the vulnerability requires local access, restricting local user permissions and monitoring for suspicious local activity may reduce risk. Users should monitor the vendor's channels for updates or patches. Patch status is not yet confirmed — check the vendor advisory for current remediation guidance.