[FEATURE] Invert plan mode "clear context" setting — show by default, hide via opt-out

Preflight Checklist

• [x] I have searched existing requests and this feature hasn't been requested yet
• [x] This is a single feature request (not multiple features)

Problem Statement

Starting from v2.1.81, the "clear context and auto-accept edits" option was removed from the plan execution prompt by default. Users must discover and set showClearContextOnPlanAccept: true in settings.json to restore it.

The current setting logic is inverted from a UX perspective:

| Current (v2.1.81+) | Proposed |
|---|---|
| Option hidden by default | Option visible by default |
| showClearContextOnPlanAccept: true to restore | hideClearContextOnPlanAccept: true to remove |
| Power users must discover a hidden flag | Power users who don't want it can opt out |

Why the Current Approach Is Problematic

1. Discoverability: The majority of users who benefit from context clearing at plan acceptance now need to find and set a configuration flag to restore a feature that was previously visible by default. This is a regression in usability — the option should be discoverable at the point of use, not buried in settings documentation.

1. Context rot at high utilisation: Even with the 1M context window, model performance degrades well before the window is full. In practice, context rot becomes noticeable at approximately 30% utilisation (~300K tokens). The plan acceptance prompt is a natural checkpoint to clear context and start execution fresh.

1. Compounding token cost: Every turn sends the entire accumulated context as input tokens. At 300K context, each subsequent exchange costs 300K+ input tokens to process, and the response adds to it further. Clearing at plan acceptance breaks this compounding cycle at exactly the right moment.

1. Plans persist to disk: Plans are saved to ~/.claude/plans/, so clearing context before execution is safe — the plan survives the clear. Without the option, the workaround is: Escape the prompt → /clear → re-invoke the plan. This is clunky and easy to forget.

1. Community confusion: Multiple issues have been filed around this change:

• #38071 — Requesting the option back
• #37688 — Users expecting auto-clearing but getting neither auto-clearing nor the manual option
• #18599 — The original complaint was about accidental context clearing (option ordering), not about removing the option entirely

Proposed Solution

Invert the setting:

• Show "Clear context and auto-accept edits" in the plan execution prompt by default
• Provide hideClearContextOnPlanAccept: true for users who want to remove it
• Optionally, reorder the options so "clear context" is not the first/default selection (addressing the original concern from #18599)

Why This Ordering Matters

The original issue (#18599) complained that "clear context" was the default selection (first option, activated by Shift+Tab), causing accidental context loss. The correct fix for that is reordering — making "auto-accept edits" the first option while keeping "clear context" visible as a secondary choice. Hiding the option entirely conflates two different problems: accidental activation (ordering) and feature availability (visibility).

Alternative Solutions

Currently using showClearContextOnPlanAccept: true in settings.json as a workaround. Without this, the flow is: Escape prompt → /clear → re-invoke the plan manually.

Priority

High - Significant impact on productivity

Feature Category

Configuration and settings

Use Case Example

1. Enter plan mode for a complex multi-file refactoring task
2. Spend 50K-100K+ tokens discussing and refining the plan with Claude
3. Plan is finalised and saved to ~/.claude/plans/
4. At plan acceptance, context is already at ~30% utilisation
5. Without "clear context" option visible, I must escape, /clear, then re-invoke
6. With the option visible (as it was before v2.1.81), it's a single decision point

Additional Context

Environment Info

• Claude Code: v2.1.83
• Plan: Claude Max
• Platforms: Linux (Kali), Windows