investigate

To find the root cause of a 500 error or stack trace, this skill drives a four-phase workflow: investigate symptoms, analyze evidence, hypothesize the cause, and implement a grounded fix only after confirming the actual issue.

Debugging a regression where code was working yesterday requires tracing the failing code path with search and reads, validating hypotheses through reproduction, and enforcing a no-fix rule until the actual root cause is identified.

Systematically investigate unexpected behavior by moving through a structured workflow from evidence to grounded fix: collect symptoms, analyze the failing code path, hypothesize the root cause, and validate it before implementing changes.

Yes, root cause analysis enforces an iron law guardrail that blocks symptom-only patches during incident response by requiring you to confirm the actual cause through reproduction and evidence tracing before implementing any fixes.

To start debugging an error, you need the exact error message, the stack trace, and the steps to reproduce the issue to produce a root-cause hypothesis and a fix plan that avoids symptom-only changes.