What problem does it solve?
Debugging often devolves into guesswork: fixes applied without root-cause analysis, repeated failed attempts, and leftover debug logs polluting the codebase. This Skill enforces a disciplined debugging methodology so bugs are reproduced, isolated, and fixed at the root cause rather than patched at the symptom level.
Core Features & Use Cases
- Feedback Loop First: Requires a reproducible failing test, HTTP script, CLI replay, or minimal verification script before any code change, ensuring the fix can be proven to work.
- Four-Phase Method: Structured locate, analyze, hypothesize, and fix stages with falsifiable hypotheses and single-variable testing.
- Git History as Memory: Checks recent commits and past reverted fixes to avoid repeating previously failed approaches.
- Escalation Ladder: After repeated failures, progressively escalates from reflection to architecture review to user consultation instead of endless blind retries.
- Use Case: A test suite fails intermittently in CI. The Skill guides the agent to reproduce the failure locally, trace the data flow, form ranked hypotheses, write a failing test, apply a minimal root-cause fix, and clean up all temporary debug logs before delivery.
Quick Start
Ask the agent to debug the failing test or unexpected behavior using the hello-debug workflow, starting with a reproducible feedback loop.
