Problem it solves
Multi-agent runs produce separate logs per agent. There's no unified view showing how the network ran as a whole, how agents related to each other, or what the full footprint was.
When to use
After every multi-agent run. This pattern is always required when the orchestration pattern is used.
When not to use
Never optional for multi-agent runs. There is no scenario in which a multi-agent run can close without an aggregate audit record.
Governing principle
The aggregate audit is the accountability record for the full network run. It cannot be reconstructed from individual agent logs after the fact — it must be composed as part of the run.
Required Components
Interaction Flow
Run completes or terminates
Whether the full run succeeded, partially failed, or was aborted, the aggregate audit is composed.
Network timeline assembled
All agent actions, handoffs, consent events, and tool calls are composed into a single chronological timeline.
Per-agent view available
Each agent's individual record is preserved and accessible via drilldown from the network view.
Cross-agent events highlighted
Handoffs, shared context, conflicting outputs, and consent events that span multiple agents are highlighted in the network view.
Export available
The full aggregate record can be exported in a structured format for compliance, incident review, or archival.
Governance requirements
Aggregate audit records must be immutable after composition. Individual agent logs and the aggregate record must be consistent — the aggregate cannot be a filtered summary.
Accessibility notes
The aggregate timeline must be navigable by keyboard. Per-agent drilldown must use aria-expanded. Export functions must not require mouse-only interactions.