What problem does it solve? When a bug report, failing test, CI failure, or production incident lands on your plate, ad-hoc guessing wastes hours and often patches symptoms instead of causes. This Skill enforces a disciplined root-cause workflow: build a fast deterministic reproduction loop, generate ranked falsifiable hypotheses, fix via TDD, and add regression coverage so the bug class never returns. ## Core Features & Use Cases - Feedback-loop construction: Ranked strategies for building a pass/fail signal, from failing tests and curl scripts to headless browsers, bisection harnesses, and differential loops. - Systematic diagnosis: Non-deterministic bug categorization (timing, environment, state, random), layer localization decision trees, and git bisect automation for regressions. - Hypothesis-driven fixing: Generate 3-5 ranked falsifiable hypotheses, instrument one variable at a time, then fix via TDD with broader regression tests and a post-mortem for production incidents. - Use Case: A CI pipeline fails intermittently on a React frontend. The Skill guides you to classify the flakiness as timing-dependent, widen the race window with artificial delays, build a deterministic failing test, bisect to the introducing commit, and land a guarded fix. ## Quick Start Investigate this failing test and find the root cause before proposing any fix.
