debugging-and-triagelisted

# Debugging And Triage Build a reliable signal, reproduce the issue, isolate the likely cause, and return evidence to the controller. Test may diagnose and classify; product code fixes stay with the owning repository unless the task card explicitly says otherwise. ## Source Skills Used - `mattpocock/skills/engineering/diagnose`: feedback loop first, reproduce, ranked falsifiable hypotheses, targeted instrumentation, fix/regression discipline, cleanup, and post-mortem. - `systematic-debugging`: reproduce, localize, trace, fix, verify, and condition-based waiting. - `triage`: classify bug, environment, out-of-scope, missing owner, or missing evidence. - Industry basis: scientific debugging and SRE symptom/cause separation. ## Wakeflow Role Use this skill for Test-scoped failures: - a controller test card failed; - a product window backfill has uncertain evidence; - a runtime/dashboard/cold-start observation is inconsistent; - a reproducer is needed before routing to a product window. ## Stop Before Guessing Do not hypothesize from vibes. First create or identify a pass/fail signal. Try these feedback-loop forms in order: 1. Targeted failing test at the right seam. 2. CLI command with fixture input and expected output. 3. HTTP/curl script against a local or test service. 4. Browser or UI script with explicit assertions. 5. Replay captured request, event, trace, or log. 6. Minimal harness around the failing path. 7. Repeated loop for flaky behavior. 8. Human-in-