review-learning-loop

To reduce false positives in code review feedback, you identify dismissed patterns and capture candidates that led the agent astray. This allows you to propose agent policy changes that tighten comment conditions for future code review workflows.

Yes, review outcome tracking applies to both GitHub and Bitbucket workflows. You submit the PR URL, provider, and a list of comment outcomes to generate learning-loop results with references, agent updates, and summary metrics.

You need to submit the PR URL, the provider (GitHub or Bitbucket), and a list of comment outcomes. The system processes these inputs to produce references-to-create, suggested agent updates, and summary metrics.

No, the review learning process applies safety constraints to avoid copying comment wording or storing personal and emotional content. It focuses strictly on generalized reference candidates and measurable metrics.

Creating generalized review references involves capturing accepted and resolved judgments from code reviews. The system transforms these outcomes into reusable team knowledge and suggests agent updates using measurable metrics.