gsd-debug

Persistent state debugging preserves investigation progress by orchestrating isolated subagents and managing session files across context resets. It stores checkpoints in .planning/debug files, allowing stepwise hypothesis testing and human verification without losing prior diagnostic state.

Systematic debugging diagnoses failing tests and CI failures by gathering expected versus actual behavior, then spawning an isolated subagent with a large context to test hypotheses. It uses guided symptom collection and automated fixes to isolate runtime errors and produce root cause reports for review.

Yes, you can start a debug session with the --diagnose flag to receive a root cause report before any fixes are applied. This mode coordinates human-verify checkpoints, letting you review the diagnosis and spawn continuation agents to apply remediation only when ready.

Subagent orchestration spawns a fresh debugger subagent with a large context window to investigate runtime errors independently. It manages persistent session state files, handles human verification checkpoints, and can spawn continuation agents to apply fixes or continue the incident response investigation.

No, the debugging workflow handles creating and reading .planning/debug session files automatically during symptom gathering. You simply describe the observed failure, and the system manages the persistent state and checkpoints required for the investigation.

Stateful debugging checkpoints are ideal for complex, multi-step investigations involving flaky CI jobs or reproducible bugs where manual reproduction is tedious. They preserve diagnostic state across context resets, enabling stepwise hypothesis testing and human verification without restarting the investigation.