Governing Autonomous AI

If a test self-heals, did it fix a flaky locator, or did it mask a real bug?

Setting Confidence Thresholds

You can configure the AI:

High Confidence (95%+): Auto-heal and pass the test.

Medium Confidence (75-94%): Warn the user but pass.

Low Confidence: Fail the test.

Human in the Loop

Quality Architects must review "Heal Reports" weekly to ensure the AI isn't adapting to broken user experiences.