The X indexed snippet presents a small but executable Codex agent configuration idea: have Sol create a custom agent named luna_worker, set its model to gpt-5.6-luna, and then have that agent take on subagent tasks. The second snippet asks you to preserve the existing configuration, validate it against the installed version, show the diff, and only then use luna_worker.
create a custom agent named luna_worker at ~/.codex/agents/luna-worker.toml. use these settings: model = "gpt-5.6-luna"… This is a necessary excerpt; read the original source for full context.
The body of the X post could not be expanded in the Tabbit international edition, so this article retains only the original text visible in Google's index. Before actually executing it, have the agent first read the current Codex documentation and configuration format, back up the file, show the diff, and confirm whether the custom agent will inherit permissions or tools it should not inherit.
GPT-5.6 Luna
