codex-first

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 1.00). The URL is high-risk: it is a direct install script piped to sh from chatgpt.com (not the known official openai/chat.openai.com host), which is an untrusted domain for distributing executables and could easily deliver malware.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly tells users to run a remote install script with curl -fsSL https://chatgpt.com/codex/install.sh | sh, which fetches and executes external code to install the required codex CLI.

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 instructs using a "--dangerously-bypass-approvals-and-sandbox" flag (and encourages --yolo / curl|sh installs and unconstrained command execution) which bypass security mechanisms and would allow arbitrary actions on the host.