Clarify and test the boundary between AI-generated plans and user-approved editing actions.\n\nAcceptance criteria:\n- Document which actions are proposal-only.\n- Add checks for destructive or unsupported operations.\n- Record security assumptions in SECURITY.md or a dedicated docs page.
Clarify and test the boundary between AI-generated plans and user-approved editing actions.\n\nAcceptance criteria:\n- Document which actions are proposal-only.\n- Add checks for destructive or unsupported operations.\n- Record security assumptions in SECURITY.md or a dedicated docs page.