Feature Request: Allow configuring default action for “Implement this plan?” (prefer staying in Plan mode)
What variant of Codex are you using?
CLI
What feature would you like to see?
Summary
In Plan Mode, Codex prompts:
"Implement this plan?"
Currently, the default action is effectively "Yes" (execute/implement) when pressing Enter.
I would like the ability to configure the default behavior to:
"No, stay in Plan mode"
---
Problem
When iterating on a plan, it's easy to accidentally press Enter and trigger execution prematurely.
This leads to:
- Unintended code changes
- Interrupted planning workflow
- Extra overhead reverting or stopping execution
For users who primarily use Plan Mode for iterative design, the current default is risky.
---
Expected Behavior
Provide a way to configure the default selection for the confirmation prompt.
For example:
- Default = "Implement" (current behavior)
- Default = "Stay in Plan mode" (requested)
---
Proposed Solutions
Any of the following would solve the issue:
- Config option (preferred)
``toml``
plan_confirmation_default = "stay"
- CLI flag / environment variable
``bash``
CODEX_PLAN_DEFAULT=stay
- Keybinding behavior change
- Enter = stay in plan
- Ctrl+Enter = implement
- Optional: remember last choice per session
---
Alternatives Considered
- Using prompts like "do not implement" → not reliable
- Avoiding Plan Mode → loses structured planning workflow
- Hooks → cannot intercept or modify this UI behavior
---
Additional Context
This feature would improve safety and usability for users who:
- Iteratively refine plans
- Prefer explicit execution triggers
- Work on large or sensitive codebases
---
Impact
Low complexity, high UX improvement:
- Reduces accidental execution
- Aligns with "safe by default" workflows
- Makes Plan Mode more practical for real-world usage
---
Thanks for considering this!
Related requests
<https://community.openai.com/t/feature-request-plan-mode-add-an-optional-message-box-when-confirming-plan-execute-in-codex-vs-code/1374051/2?utm_source=chatgpt.com>
Additional information
_No response_
This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗