debugging-empty-output

AI workflow sessions complete with no usable output when agents fail to call the required complete() tool to emit final output, or when prompts incorrectly instruct the LLM to stream prose instead of capturing structured results.

Troubleshoot an empty artifact list by following an ordered diagnostic checklist that sequentially reviews session AI summaries and inspects agent tool call logs to pinpoint the root cause of the missing output.

The "Session completed without producing a summarizable document" message is a sentinel indicating the workflow finished without emitting final output, typically because the agent never called the required complete() tool.

Avoid common incorrect fixes when debugging empty AI output such as migrating workflow execution patterns or adding redundant platform tools, which waste troubleshooting time without addressing the actual missing complete() tool call.

Yes, agent prompts must explicitly instruct the LLM to call the required complete() tool to capture structured results, rather than streaming prose, to prevent empty summaries and ensure artifacts are properly emitted.

First check the session AI summaries and agent tool call logs to determine whether the complete() tool was invoked, as this directly identifies why the outputTo action failed to capture final output.