Run Codex CLI for $ARGUMENTS.

Defaults

Primary review command: codex review
Primary general command: codex exec
Default model: gpt-5.4 (referred to as $MODEL in recipes below)
Default review / consult reasoning config: -c model_reasoning_effort=xhigh
Default research reasoning config: -c model_reasoning_effort=high
Default research search path: top-level --search
Default posture: review-only first, write only when the user explicitly wants Codex to apply changes
Preferred automation path for writes: --full-auto before --dangerously-bypass-approvals-and-sandbox
Keep sandbox bypass available for exceptional cases, but only after explicit user confirmation for that specific run
Resume path: codex exec resume <session_id> "<follow-up>"

codex