Why should an environment distinguish seeing a new rule from being authorised to apply it?

**If an agent can read the policy, should it be able to update every affected record?**

No. Access to guidance does not establish authority to change a decision. In insurance operations, a claims handler may read a revised settlement rule while only a delegated reviewer can reopen an authorised settlement. The handler can identify the issue and submit a review request without changing the settlement directly.

**What should the environment do with an unauthorised attempt?**

It should preserve the protected state, return a meaningful permission result and expose an allowed next action. Silently accepting the update would teach the agent that policy correctness overrides operational controls. Rejecting it without a usable escalation path would make the task impossible for reasons unrelated to policy reasoning.

**What could MirrorFirm generate?**

A generated environment can contain role assignments, delegated authority limits, restricted transitions and review requests linked to affected records. Builders can use these elements to teach an agent when to act, when to prepare evidence and when to transfer a decision. This makes policy propagation a coordinated activity with accountable authority.