CVE-2026-79408

MetaGPT 0.8.1 contains an OS command injection flaw (CWE-78) in the path argument of RepoParser.rebuild_class_views() in metagpt/repo_parser.py. An attacker who can influence the value passed as path to this function can have arbitrary operating system commands executed on the host running MetaGPT. Per its CVSS 3.1 vector (AV:N/AC:L/PR:N/UI:N), the flaw is rated as exploitable over the network without authentication or user interaction, yielding high impact on confidentiality, integrity, and availability. Users running MetaGPT 0.8.1 are affected, particularly those processing attacker-supplied repository paths or integrating the repo parser into automated or network-reachable workflows. No public proof-of-concept, KEV listing, or confirmed exploitation has been reported; EPSS currently estimates a 1.2% probability of exploitation within 30 days.

What to do: Upgrade MetaGPT from 0.8.1 to a patched release as soon as one is published and monitor the upstream project's advisories for the fixed version. Until then, avoid passing untrusted or user-controlled values as the path argument to RepoParser.rebuild_class_views(), validate/sanitize that input, and run MetaGPT on untrusted repositories only in sandboxed environments. Audit integrations, scripts, and CI pipelines that invoke metagpt/repo_parser.py with external input to confirm nothing network-reachable feeds attacker-controlled paths into it.