How should scheduled reviews enter an agent environment?

A simulated environment can expose a clock and a review scheduler alongside business records. A schedule record could contain `schedule_id`, `subject_id`, `cadence_rule`, `timezone`, `next_due_at` and `schedule_status`. A separate obligation record links the resulting review to the schedule that caused it.

The distinction matters because a calendar rule is not itself a work item. When `advance_clock` crosses a due boundary, the scheduler can emit a `review_due` event and create an obligation. Reprocessing the same boundary should not create duplicate work; a stable trigger key can identify that occurrence.

For an Australian utility account, the supplied timezone determines which local date makes the review due. The model should inspect the obligation through actions such as `list_due_reviews` and `get_review_context`, rather than receive all relevant facts in its opening prompt.

MirrorFirm could scope this as a generated environment with a fixed starting clock. An evaluation hook can record whether the agent discovers the obligation before taking unrelated actions, making time-driven task recognition directly observable.