What problem does it solve? When something appears broken, teams often rewrite working systems based on guesses. This Skill enforces evidence-based root-cause investigation so failures are localized to the correct layer before any change is made, protecting working state and preventing shotgun fixes. ## Core Features & Use Cases - Layer Classification: Separates symptoms into execution, persistence, data, interpretation, presentation, environment, dependency, permission, or expectation-mismatch domains. - Hypothesis Tracking: Maintains a table of hypotheses with evidence for/against, discriminating tests, and confidence, retiring disproven theories. - Protected-State Discipline: Declares what must not change during diagnosis and prefers read-only evidence gathering first. - Structured Decisions: Returns explicit outcomes such as CAUSE SUPPORTED, LAYER LOCALIZED, NO FAILURE REPRODUCED, EXPECTATION MISMATCH, BLOCKED, or STOP. - Use Case: A UI banner stays blank after a scan. Instead of rewriting the scanner, the investigation shows records were written correctly, localizing the failure to the presentation layer before any bounded fix. ## Quick Start Ask the AI to investigate why a reported failure occurred, localizing the failed layer and cause with evidence before proposing any fix.
