Plan execution prompt lost 'clear context and execute' option
Description
The plan execution prompt used to offer an option to clear context before executing. This was removed in a recent update, and the prompt now only shows three options:
- Yes, auto-accept edits
- Yes, manually approve edits
- Type here to tell Claude what to change
Previously there was a fourth option to clear context and execute, which was valuable because plan execution is exactly when context tends to be high — you've just spent tokens discussing and refining the plan.
Why this matters
Plans persist to ~/.claude/plans/, so clearing context before execution is safe and often desirable. Without this option, the workaround is:
- Escape the prompt
/clear- Re-invoke the plan
This is clunky and easy to forget, especially when context is already near the limit. The old flow was a single decision point: "yes, and start fresh."
Context visibility
Separately, the context usage percentage / bar is no longer visible at the plan execution prompt. I have a custom statusLine that renders context_window.used_percentage, but at the plan confirmation screen the status line isn't shown. Seeing context level at the moment you're deciding whether to execute is when it matters most.
Expected behavior
- Plan execution prompt includes a "Clear context and execute" option (or similar)
- Context usage is visible at the plan confirmation screen
Version
Claude Code 2.1.81
Screenshot
The current 3-option prompt:
_(Three options only — no clear-and-execute, no context level visible)_
11 Comments
I had the same problem! I tried it on
2.1.80, and it seems to have returned. I suggest, until Anthropic fixes this problem in later patchs, to use2.1.80.<img width="943" height="654" alt="Image" src="https://github.com/user-attachments/assets/cbf5bc8b-f0a5-44d7-b4c1-7d8cf555db2e" />
This was unshipped because you don't need this anymore with 1M context! To opt back into it, ask Claude to set
showClearContextOnPlanAccept: truein your settings.json. See docs: https://code.claude.com/docs/en/settings.@bcherny I think
showClearContextOnPlanAccept: trueshould be the default with themodel: opusplan.I agree! This was such an unprecedented change, that if your own userbase has to open up an Issue about it, you've done something wrong. You did NOT cook, Anthropic.
It has less to do with the limits of the context window (1M window is great!) and more to do with wanting to avoid wasting tokens unnecessarily, thus hitting session limits faster.
I appreciate there is an opt-in. While 1M context can be very useful, I observe myself getting better results with a great plan and empty context.
Though I wonder if a compact rather than a clear would do even better.
Until there is clear testing that proves that models perform the same at 700k token context as 30k token, this statement isn't valid. The needle-in-the-haystack benchmarking showing claude is ~70% accurate with long contexts just hasn't been my experience. Its still far worse (after 300k-400k tokens).
Even if it were the case, this just increases the token spend for the users. It's not a good experience.
Yes, please give us back the option to clear context.
from very much use it seems to me that higher context causes worse results. this is quite dismissive.
This is absolutely right !
This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.