review

Automate quality assurance by triggering a review process that detects critical research items like completed stages or draft papers, invoking a judge agent to assess results against plans. This ensures pipeline stages meet predefined standards before proceeding.

Yes, you can audit plan versus execution deviations using the audit mode. The review skill evaluates updated experimental results and completed stages, identifying discrepancies to provide a pass or fail verdict for the research project.

Supported review modes include 'paper' for evaluating draft papers, 'stage S{N}' for validating specific research stages, and 'audit' for comparing planned tasks against actual execution outcomes within the research pipeline.

Intelligent target detection works by automatically identifying the most critical item requiring review based on project status, such as completed experimental stages or updated results, then invoking the judge agent to assess the targeted research output.

Yes, the review skill supports user-defined focus queries to guide the paper review process. These queries allow you to direct the judge agent's attention to specific quality assurance criteria during the automated assessment.

When a critical failure is detected during a research stage review, the skill manages review rounds, persists the review outcomes, and escalates the critical failure to ensure the research pipeline meets predefined quality standards.