investigate

Root-cause debugging is a systematic investigation process that identifies the underlying cause of bugs or performance regressions before applying fixes. You need it when understanding the true source of unexpected behavior matters more than patching symptoms.

To reproduce and analyze a crash without chasing symptoms, follow a strict four-phase workflow: investigate, analyze, hypothesize, and implement. This method gathers evidence and validates testable hypotheses to ensure you fix the actual root cause.

Yes, you can use shell commands to gather evidence during root-cause analysis. The investigation process interacts with shell commands and system state via allowed-tools to safely reproduce bugs, inspect system conditions, and verify findings auditably.

The best way to fix a performance regression safely is to validate testable hypotheses and verify fixes before applying changes. This ensures the regression's root cause is confirmed and the resolution is auditable, preventing new bugs.

Your bug fix keeps failing because you are patching symptoms instead of the root cause. Without a structured investigation to generate and test hypotheses, the underlying fault remains, causing the error to persist or reappear.