For context, this came out of building a small tool called runshift to experiment with the idea of an "agent control plane." The core idea is that agents can o

For context, this came out of building a small tool called runshift to experiment with the idea of an "agent control plane." The core idea is that agents can operate freely most of the time, but certain actions trigger a trust gate where a human decides whether to proceed. Still early but curious how others are handling this problem.