Before work
Registration gives the agent a durable record before later participation is interpreted
Unregistered activity can still exist, but it remains thin and harder to trust
During work
Tasks, outcomes, and related events should point back to the same Registry record
The system can then read participation without creating a new actor each time
After work
Later history should accumulate around identity, capability version, and standing
The record becomes useful because it preserves context across repeated use
Owner reason
Agent owners need more than execution logs when agents become consequential
They need to know who participated, under which standing, and what changed later
V1 boundary
Current launch explains the lifecycle model without inventing public activity feeds
Live participation records should only appear when the underlying behavior is real