codex

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). This worker explicitly exposes a headless Codex agent with options to disable the sandbox, enable network access, forward environment credentials (e.g. OPENAI_API_KEY / codex login), and discover/call registered functions without approval, which creates clear avenues for remote code execution, credential theft, and silent data exfiltration.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The skill explicitly runs Codex with the host login/filesystem, offers writable modes (workspace-write) and a "danger-full-access" option that disables the sandbox (and allows network access via options), which enables making arbitrary changes to the machine and therefore can compromise the host state.