OpenAI patched a command injection flaw in its Codex CLI tool that let attackers run arbitrary commands on developer machines by hiding malicious configuration ...