DataInterpreter in MetaGPT’s data_interpreter.py can be abused to inject code when invoked locally. The flaw allows an attacker who can trigger the function to execute arbitrary statements, potentially compromising local processes or leaking sensitive information. The vulnerability is a typical code-injection weakness identified by CWE‑74 and CWE‑94.

FoundationAgents’ MetaGPT versions up to 0.8.2 are affected. The issue resides in the DataInterpreter routine supplied with those releases.

The CVSS score of 4.8 signals a moderate severity risk, and the absence of an EPSS rating indicates limited known exploitation activity. Because the attack vector is strictly local, the threat is confined to systems where the user or process can invoke the vulnerable function. The vulnerability is not listed in CISA KEV, but the public exploit code suggests that an unaware user could leverage it if local access is available.