io.github.ipezygj

Safety checks an agent runs before it acts: package malware, destructive command, secret leak.

Statistical checks an agent runs before trusting an AI eval number (is #1 real, judge bias, more).