What problem does it solve? Debugging sessions often lose progress when AI context windows reset, forcing you to re-explain symptoms and re-run investigations. This Skill orchestrates systematic debugging with persistent checkpoint state, so investigations survive context resets and can be resumed at any time. ## Core Features & Use Cases - Scientific Method Debugging: Spawns an isolated gsd-debugger subagent that investigates bugs systematically rather than guessing at fixes. - Persistent Session Management: Supports list, status <slug>, and continue <slug> subcommands to track and resume debug sessions stored under .planning/debug/. - Diagnose-Only Mode: The --diagnose flag returns a Root Cause Report without applying any fix, useful when you want analysis before changes. - Use Case: You hit a failing test in a large codebase. Invoke the skill with the symptom description, let the debugger agent isolate the root cause, then continue the session the next day without losing any investigation state. ## Quick Start Invoke $gsd-debug followed by a description of the bug symptoms to start a systematic debugging session.
