Under the hoodTier: how much it should matter to you
2Useful: my rating, 1 to 5
2Signal: worth watching, 1 to 5
AgentsArea: what it touches
Internal ChangesKind: in v2.1.293,
Internal ChangesSection of the release
Unclear The name of the new field is not shown, nor whether Claude Code actually sends it yet.
What
The Agent SDK's output format, which programs that drive Claude Code read, has a new description for one kind of run. It covers a subagent (a helper agent Claude starts to handle part of a task) that is resumed by a tool call, where the tool call waits for it to finish. SendMessage works this way when background tasks are turned off.
For such a run:
the run has ended once the resuming tool call returns or fails
the subagent's report is placed in that tool call's result
the run's task_started event names the tool call in its tool_use_id
The description is marked as internal.
Why
Programs built on the SDK get a way to recognise these resumed runs and know that the waiting tool call's result holds their report.
How sure we are
Something disagreesSomething we can check disagrees with this entry, or the writer said they could not settle it.
The writer flagged doubtThe name of the new field is not shown, nor whether Claude Code actually sends it yet.
Anthropic's release notes agreeFixed Claude being told to continue or message subagents with SendMessage in sessions, including resumed ones, where a host, a permission…
