Skip to contents

How one delegation ended. The runtime's record (IDs, status, stop_reason) is kept apart from what the subagent's model said (answer and claims). A "completed" status means the run ended normally, not that the task succeeded. LeadAgent returns outcomes from $parallel_delegate() and, as JSON, from the delegate_to_agent tool. Fields are read with $.

Usage

DelegationOutcome(runtime, answer = "", references = list(), claims = list())

Arguments

runtime

List of IDs, status, stop_reason and other facts recorded by the runtime.

answer

The subagent's final reply, cut to 8 KiB. A longer reply is saved in full in references.

references

Saved artifacts, such as the full answer or large tool results, with their origin and whether they are still available.

claims

The subagent's report of missing_evidence and unresolved_work from its structured output. NULL means it didn't report, not that nothing is missing.

Value

A DelegationOutcome object.