Replay resistance is the set of controls that stop a previously valid request, token, or approval from being used again later. A system with replay resistance checks freshness and context so an old message cannot be reused as if it were new authority.
This matters because attackers often capture or delay legitimate requests and try to resend them to repeat a privileged action, such as approving access, revoking sessions, or triggering a deployment. Good defenses use nonces, short lifetimes, request binding, timestamps, one-time tokens, and state checks at execution time. In agent workflows, replay resistance helps ensure that the approval still matches the exact user, action, resource, and moment before the system acts.