Perf Analyzer

It takes your baseline measurements, experiment results, and hypothesis tests, then outputs a structured report with evidence-backed recommendations, failed experiments worth abandoning, and clear next steps. The constraint that matters most is it only cites evidence that actually exists in your logs or code. If your data is insufficient, it tells you to re-run rather than speculating. Useful when you need to move from "here's what we observed" to "here's what we should do about it" without losing the thread of what you actually proved.