codex-prompting

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The skill content intentionally configures an autonomous Codex agent with "approval_policy = 'never'" and "sandbox_mode = 'danger-full-access'", plus tooling for arbitrary shell commands, apply_patch, exec/PTY access, and escalated permissions, which collectively enable remote code execution, data access/exfiltration, and silent code modifications (supply‑chain/backdoor risk) without human oversight.