I wrote a follow-up to my earlier “Codex skills as RE playbooks” post. This time I ran the same two RE skills across OpenAI Codex vs Claude Code with a static-first workflow and explicit execution gates.

Main takeaways: Codex felt more autonomous for driving the workflow and producing strict artifacts. Claude produced a stronger “analyst report” output (clearer narrative, gaps, and next steps).