I built Ctrl, an open-source execution control plane that sits between an agent and its tools. Instead of letting tool calls execute directly, Ctrl intercepts them, dynamically scores risk, applies policy (allow / deny / approve), and only then execute…