Why Agent Approval Must Happen Before the Tool Call Executes

A practical operator-gate architecture for pausing risky AI agent tool calls and enforcing human approval before side effects occur.