What problem does it solve? When a test fails or a build reports blocked, developers often patch symptoms blindly instead of finding the root cause. This Skill enforces a diagnose-first workflow that reproduces the issue, ladders hypotheses, and applies only evidence-backed minimal fixes. ## Core Features & Use Cases - Root Cause Diagnosis: Spawns a single debugger subagent that reproduces the failure and isolates the cause with evidence before touching code. - Hypothesis Laddering: Systematically works through candidate causes and returns the full chain when it cannot fix the issue outright. - Outcome Routing: Routes results by verdict — fixed issues go to code review, stuck cases surface the hypothesis ladder for user decision, and non-reproducible bugs request exact repro steps instead of guess-patching. - Use Case: A test suite fails after a feature build. Invoke the Skill with the failing test name, and it returns the root cause, supporting evidence, the minimal fix, and the re-run output. ## Quick Start Ask the AI to run /code-debug with the failing test name or error message to find the root cause and apply a minimal fix.
