Codex config.toml is outdated and Codex now has multi-agent support

[sla-te]

Summary

The Codex config.toml in this repo is outdated and does not match current Codex configuration behavior.

Also, Codex now has multi-agent support, and the Codex config/docs should reflect that.

What is outdated

1. approval_policy and sandbox_mode are under [permissions], but current Codex expects them as top-level settings.
2. history.persistent_instructions is legacy compared to current guidance (model_instructions_file and project instructions files).
3. Notifications use legacy [notify] table form (command, on_complete) instead of notify = ["cmd", "arg1", ...].
4. [features].web_search_request = true is legacy compared with top-level web_search mode.
5. Sandbox comments reference old values (off, workspace-read) instead of current values (read-only, workspace-write, danger-full-access).

Request

Please update the Codex config/docs to current schema and explicitly reflect multi-agent support.

References

• Codex config reference: https://developers.openai.com/codex/config-reference
• Codex multi-agent docs: https://developers.openai.com/codex/multi-agent