What problem does it solve?
This skill provides a structured, persistent debugging workflow that preserves investigation state across context resets so long-running diagnoses do not lose progress or context. It reduces repetitive symptom gathering and context loss when investigations require spawning fresh, high-context subagents for deep analysis.
Core Features & Use Cases
- Orchestrator workflow that checks for active sessions, gathers symptoms interactively, and either resumes or starts new investigations.
- Spawns isolated gsd-debugger subagents with a fresh large-context model to perform diagnosis and fixes while preserving the main interaction thread.
- Checkpoint and continuation handling: supports human verification checkpoints, continuation agents, and session files stored under .planning/debug/*.md for reproducibility and audit.
- Use Case: A developer reports intermittent production errors; the skill collects expected vs actual behavior, error messages and reproduction steps, then spawns a gsd-debugger agent to find and propose fixes while tracking evidence and checkpoints.
Quick Start
Start a new investigation by telling the skill the issue in one sentence, for example: Investigate a pipeline job failing with a timeout and provide recent error output.
