small ML to guide the tool

[Zierax]

you can add small ML to guide the tool like
tool will discover the target
then tool will detect interesting parameters like ?id=
then tool send to the parameter some requests by curl for get status code 500
then use automation like ghuari for exploit it

[kleozzy]

I was thinking about this as well but you really need an LLM spesifically rained for security testing as well as somewhere to host it and run it (which would require trmendous ammounts of GPU power). It will also make the process significantly slower as well.

Unless some cheap hosted security trained models arise , this seems extreme innefficient at the moment.