What problem does it solve? Temporal alerts like "sync is stale", "job timed out", or "pipeline wedged" invite immediate structural fixes — raising timeouts, splitting steps, rewriting wrappers — that are often wrong because nobody measured the actual step. This Skill enforces a measure-first discipline so you kill the theory with a stopwatch instead of a code change. ## Core Features & Use Cases - Stopwatch-first triage: Times the exact entity named in the alert (e.g. time gbrain sync --source source-a --no-embed) rather than the aggregate, so the real bottleneck is isolated. - Threshold mismatch detection: Compares the monitor's act-line against the authoritative gbrain doctor warn/fail lines to catch false pages on healthy systems. - Verdict-driven fixes: Produces a structured measurement verdict distinguishing "needs more time" from "is wedged" — two states with opposite fixes. - Use Case: A cron monitor pages repeatedly that two sources are 14 hours stale. Instead of rewriting the sync wrapper, you time the sync directly, find it completes in seconds with "Already up to date", and fix the alert threshold in two lines instead. ## Quick Start Ask the agent to triage the freshness alert by measuring the sync step before changing any timeout or threshold.
