What problem does it solve? Debugging often devolves into speculative code reading and untested fixes. This Skill enforces a disciplined diagnosis loop that reproduces the exact symptom with a deterministic failing test before any fix is attempted, ensuring the true root cause is found and verified. ## Core Features & Use Cases - Deterministic Red Feedback Loop: Builds one fast (<2s) failing command—TUnit test, HTTP request, or throwaway harness—that goes red on the bug and green when fixed. - Ranked Falsifiable Hypotheses: Requires 3-5 hypotheses with concrete testable predictions before any probing begins. - Knowledge-Graph Tracing: Uses callers/callees queries and affected-flow analysis to trace execution paths, with tagged [DEBUG-...] logs guaranteeing clean removal. - Use Case: A MediatR handler returns wrong results in production. The Skill guides you to write a failing regression test at the handler seam, minimise inputs, test ranked hypotheses, apply the minimal fix, and verify the original scenario passes. ## Quick Start Ask the AI to debug the failing checkout flow using the debug-issue workflow, starting by building a failing test that reproduces the exact reported symptom.
