Feature request: per-mode customization for Plan-mode question clarity (without forking)

Resolved 💬 6 comments Opened Feb 17, 2026 by JayceeB1 Closed May 31, 2026
💡 Likely answer: A maintainer (etraut-openai, contributor) responded on this thread — see the highlighted reply below.

What variant of Codex are you using?

CLI + IDE Extension (collaboration modes)

What feature would you like to see?

Today, Plan mode question behavior is mostly controlled by built-in collaboration mode templates.
To enforce consistently clear questions, we currently need to fork Codex or inject overrides from a custom orchestrator.

I’d like a first-class way to customize question style per collaboration mode (especially Plan mode) from user/project config, without patching source.

Use case:

  • Before asking a question, Codex should present:

1) context,
2) impact,
3) 2-4 concrete options with tradeoffs + examples,
4) recommended option,
5) one final closed question.

  • No implicit auto-advance when user has not explicitly answered (human gate remains required).

Requested capability:

  • Allow per-mode instruction override from config (e.g. plan/default), ideally file-based.
  • Stable behavior across CLI/App/IDE Extension.
  • Should affect both direct clarification questions and request_user_input flows.
  • Should not require modifying built-in template files in source.

Why this matters:
AGENTS.md/user instructions are not always sufficient because collaboration-mode developer instructions can take precedence. A supported per-mode override would remove the need for forks and improve UX consistency.

Additional information

Additional information
Related implementation points in repository:

  • core/templates/collaboration_mode/plan.md
  • core/src/models_manager/collaboration_mode_presets.rs
  • app-server turn/start collaboration_mode developer_instructions handling

View original on GitHub ↗

6 Comments

spider-yamet · 5 months ago

Hope I can contribute by resolving this issue.

JayceeB1 · 5 months ago
Hope I can contribute by resolving this issue.

this is a quite simple change imho

hope this will be take in consideration

JayceeB1 · 4 months ago

bump :)

etraut-openai contributor · 4 months ago

@JayceeB1, we generally prioritize features based on community upvotes. Bumping it won't increase the upvote count. :)

JayceeB1 · 4 months ago
@JayceeB1, we generally prioritize features based on community upvotes. Bumping it won't increase the upvote count. :)

Sorry, was worth a try 🙏🏼

etraut-openai contributor · 1 month ago

This feature request hasn't received enough upvotes, so I'm closing.