{
  "checklist_item": {
    "id": "chk_sandbox_notify_on_escape",
    "item": "If Claude tries to access something outside the sandbox boundary, the user is notified immediately and can choose whether to allow it -- the boundary lets Claude act freely inside it while still surfacing any attempted escape.",
    "category": "sandboxing",
    "source_title": "Making Claude Code more secure and autonomous with sandboxing",
    "source_url": "https://www.anthropic.com/engineering/claude-code-sandboxing",
    "summary": "Escape attempts are surfaced immediately for a human decision, rather than either silently blocked or silently allowed.",
    "created_at": "2026-07-02 06:02:48",
    "cite_as": "https://subagentchecklists.com/api/checklist-items/chk_sandbox_notify_on_escape"
  }
}