The browser-agent control plane: scope, audit, handoff, rollback

Most browser-agent demos still ask a capability question: Can the agent use the browser? That question made sense early on. A browser is messy. Pages change. Login flows break. Buttons move. If an agent can navigate a live page, recover from a bad click, and complete a real workflow, that is progress. But capability is no longer the only interesting question. The harder question is authority: Should this agent be allowed to take this action in this logged-in browser session, for this task, witho