What problem does it solve? Fixing bugs without a systematic method leads to shotgun edits, unfixed root causes, and regressions. This Skill enforces a scientific debugging loop: reproduce the bug first, lock it with a failing automated test, iterate ranked hypotheses with targeted instrumentation, then apply a minimal verified fix. ## Core Features & Use Cases - Repro Automation Lock: Blocks any code edit until the bug is reproduced as a failing automated test (Vitest, Playwright, pytest), creating a deterministic regression lock that must flip green after the fix. - Hypothesis-Driven Debug Loop: Ranks 3 candidate root causes, then iterates hypothesize → instrument → reproduce → analyze (max 5 iterations) with mandatory per-iteration evidence logged to a session file. - Minimal Fix + Handoff Report: Applies only the lines that fix the root cause, forbids drive-by refactoring, and produces a before/after proof-of-work report with manual verification steps. - Use Case: A user reports that a refund succeeds but the row status never updates. The Skill asks for expected behavior and reproduction steps, writes a failing integration test, instruments the refund flow, confirms the root cause hypothesis, applies the minimal fix, and delivers a verification guide. ## Quick Start Describe the bug with its expected behavior, actual behavior, and numbered reproduction steps, then ask the agent to debug and fix it using the scientific debug loop.
