The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/codex_delegate.py executes the local codex binary using subprocess.run. It passes arguments as a list rather than a shell string, which is a standard security measure to mitigate shell injection attacks.
[COMMAND_EXECUTION]: The skill enforces restricted execution modes by hardcoding --ask-for-approval never and selecting between --sandbox read-only or --sandbox workspace-write based on the requested task.
[COMMAND_EXECUTION]: A security check is implemented to prevent the injection of dangerous Codex flags. The script scans the prompt, model override, and worktree name for the string --dangerously-bypass-approvals-and-sandbox and terminates if detected.

codex-cli-delegate