Security alert triage with evidence-backed noise suppression

I would pay @github if they could run an LLM against all these alerts and auto-dismiss, because 99% are noise (which in of itself is dangerous because it trains people to dismiss by default, just like an ad or popup).