GitHub Security Lab makes fuzzing an agent feedback loop
The article remains in its original language.