I have been thinking about phishing simulation, how many variety of attack an organization can simulate to match real world attacks - to phish and to train, and the number is huge, no one does it. So, I have built an agent to consider social data and create a phishing attack a person might falls into. It's a very early version of idea but will love to hear thoughts.