What problem does it solve? When a bug, flaky test, or performance regression appears, it is tempting to apply speculative fixes that mask symptoms instead of removing the cause. This Skill enforces a disciplined debugging workflow that preserves evidence, reproduces the failure, tests ranked hypotheses, and only proposes a fix after the root cause is proven. ## Core Features & Use Cases - Evidence-First Investigation: Captures the observed behavior, environment, error signatures, and smallest safe reproduction before touching any code. - Ranked Hypothesis Testing: Forms one hypothesis at a time with explicit predictions, discriminating experiments, and approval gates, stopping after three failed hypotheses to reassess. - Structured Debug Report: Produces a standardized report covering symptom, reproduction, localization, root cause, smallest fix, and regression plan. - Use Case: A CI test fails intermittently. Instead of adding retries or sleeps, use this Skill to identify the meaningful state transition, record the reproduction rate, trace the failing state across boundaries, and propose a minimal root-cause fix with a regression test. ## Quick Start Ask the assistant to systematically debug the failing test or error you are seeing, find the root cause with evidence, and report it before making any code changes.