Problem

In Claude Code, plan approval has multiple variants:

1. Yes (Enter) — approve and continue
2. Yes, clear context and auto-accept edits (Shift+Tab) — approve, compact/clear conversation context, and set permission mode to auto-accept edits

Plannotator currently supports permission mode switching (acceptEdits, bypassPermissions, default) via updatedPermissions → setMode in the hook response. This covers the "auto-accept edits" part. However, there is no mechanism to trigger conversation context clearing — the "clear context" half of the shift+tab behavior.

Current behavior

When a user approves a plan in the Plannotator UI, the hook outputs:

{
  "hookSpecificOutput": {
    "hookEventName": "PermissionRequest",
    "decision": {
      "behavior": "allow",
      "updatedPermissions": [{ "type": "setMode", "mode": "acceptEdits", "destination": "session" }]
    }
  }
}

This restores the permission mode but does not signal context clearing.

Desired behavior

Plannotator should offer an approval option equivalent to shift+tab — "Approve & clear context (auto-accept edits)" — that both:

1. Sets the permission mode to acceptEdits (already supported)
2. Triggers conversation context clearing/compaction

Open question

It's unclear whether Claude Code's hook response API currently supports a clearContext signal. If not, this may also require a feature request upstream to Claude Code's hook API to expose this capability. If the API does support it already (or gains support), Plannotator could add it as a checkbox or button variant in the approve UI.

Context

The shift+tab approval variant is valuable for long planning conversations where the plan itself captures all needed context — clearing the conversation before implementation reduces noise and token usage.