For the next several days, we’re going to ask our AI agents to do a bit of role playing. What I’ve found is that AI has a hard time being an absolute expert on everything all of the time. But with some focused prompting, it can be an expert on the topic you need right now.

Today, we’re starting with security. Those that are both pro- and anti-vibe coding will agree that one of the aspects of AI-assisted software development that falls apart quickly is locking your systems down appropriately, to keep out the bad actors.

Using your agent to also evaluate their own work goes a long way towards success.