Intel

AIKIDO-2026-11020

@openai/codex is vulnerable to Protection Mechanism Failure

Protection Mechanism Failure Pre-CVE
Found by Aikido Intel before public disclosure or CVE publication.
Published Yesterday

55

Medium Risk

This Affects:

JS@openai/codex
0.0.0 - 0.135.0
Fixed in 0.136.0
Are you affected? Scan for Free

TL;DR

A vulnerability in command-safety classification and local WebSocket handling could allow unintended code execution or unauthorized local connections under specific conditions. On macOS/Linux, PowerShell-looking commands were incorrectly routed through the Windows PowerShell parser, which could execute a repository-controlled pwsh binary during safety validation, before the normal sandboxed execution path. An attacker might exploit this by placing a malicious pwsh executable in a controlled path and tricking the system into classifying it as safe, or by initiating browser-originated requests to the local exec-server WebSocket listener to reach functionality that should only be accessible to trusted local clients.

Who does this affect?

You are affected if you are using a version that falls within the vulnerable range.

Background info

@openai/codex is vulnerable to Protection Mechanism Failure in versions 0.0.0 - 0.135.0.

How to fix this

Upgrade the @openai/codex library to the patch version.