What problem does it solve? When a project's tests fail, an end-to-end verification request breaks, or a user reports a bug in an already-built app, this Skill keeps debugging until the failing behavior actually works instead of stopping after reporting the error once. ## Core Features & Use Cases - Reproduce-first debugging: Always re-runs the failing test, endpoint, or user flow to capture the real traceback before editing any code. - Structured diagnosis: Checks cheapest causes first, including missing dependencies, API contract drift between frontend and backend, framework API mismatches (CrewAI, DSPy, MCP, GraphRAG), stale environments, and missing data files. - One-hypothesis iteration: Changes one thing at a time and re-runs the reproduction immediately, repeating until the failure is resolved or the user must decide something. - Use Case: A full-stack app's test suite fails after a backend route change; the Skill diffs the frontend call against the API contract in design.md, fixes the deviating side, and re-runs the full test gate plus end-to-end verification. ## Quick Start Ask the agent to debug the failing test in the current project and keep iterating until the test suite and end-to-end verification pass.
