autoresearch

To automate a research loop with validation, you need a stateful workflow that continuously nudges research until validation evidence is present and gates completion on passing validation, ensuring outputs are verified rather than just model-signaled.

Artifact-gated completion in a stateful research workflow ensures the loop only finishes when validation evidence is present in the output artifacts, rather than stopping based on arbitrary model signals.

To set up persistent state between automated research runs, use a native-hook persistent research loop that maintains state across executions and continuously validates research outcomes until completion criteria are met.

Yes, research automation supports selecting between script validation and prompt plus architect validation modes, allowing you to choose the validation approach that best fits your workflow and evidence requirements.

To start an automated research loop, initiate intake with the deep-interview command, then execute the research loop using the autoresearch command to trigger the stateful validation workflow.

An automated research workflow finishes without validating results if it lacks artifact-gated completion. Without a stateful loop enforcing validation evidence, the workflow relies on model signals instead of verified outputs.