What problem does it solve? Debugging often stalls on guesswork, unverified assumptions, and fixes that never get validated against the original symptom. This Skill structures bug and performance investigations so every hypothesis is testable, every fix is confirmed against the reported behavior, and intermittent failures are measured rather than chased blindly. ## Core Features & Use Cases - Hypothesis-driven investigation: Trace symptoms through actual callers, logs, domain contracts, and recent changes, testing the cheapest discriminating observation first. - Regression and performance checks: Build the smallest practical check (test, request, CLI invocation, UI flow, or replay) that exercises the reported behavior, with baselines for performance work. - Human-in-the-loop reproduction: Adapt the included scripts/hitl-loop.template.sh capture template when reproduction requires manual interaction, recording triggers and reproduction rates. - Use Case: A user reports that clicking Export intermittently throws an error. Use this Skill to trace the callers, form falsifiable hypotheses, run the HITL capture script to record the error, fix the shared root cause, and re-run the original scenario with a focused regression check. ## Quick Start Use the diagnosing-bugs skill to investigate why the export button intermittently throws an error and confirm the fix with a regression check.
