Point AI at your own work before someone else does (building securely with AI / tips)

Yeah, this reminds me of old-school second-order attacks like second-order SQLi or even blind XSS. Good approach and definitely needs to be in peoples threat models.