Allow configuration of AI agents to ask clarifying questions

I used to say AI coding agents are like junior engineers. I decided that was wrong. Here's why. Yes, coding agents pretend to know everything. They never ask for help. They don't ask "what's the architecture?" or "should this be reusable?" They can loop endlessly on the same debug approach without stepping back. I've seen these traits in junior engineers over the years, including myself back in the day. So the comparison seemed apt. But here's what I missed: junior engineers grow. They develop creativity, humility, and fresh perspectives that challenge assumptions. The best ones eventually ask the questions I forgot to ask myself. AI doesn't do that. 𝗡𝗼𝘁 𝘆𝗲𝘁, 𝗮𝗻𝘆𝘄𝗮𝘆. That said, AI coding agents are a genuine shift in how software gets built. The leverage is real. But only if we stop treating them like juniors we're waiting to grow up. Teams that make that mistake? They ship hallucinated logic to production. They burn months chasing phantom productivity. They lose credibility with stakeholders who were promised acceleration and got cleanup work instead. This isn't anecdotal: • Trust in AI accuracy has fallen from 40% to 29%—even as usage climbed to 80%. • Only 17% of developers say AI agents improved team collaboration. The gains aren't scaling. The gap between usage and trust isn't shrinking. This is a skills problem, not a waiting problem. The skills that unlock AI's real potential are the same ones we've always valued: • 𝗔𝘀𝗸𝗶𝗻𝗴 𝘁𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻𝘀 others didn't think to ask • 𝗙𝗼𝗿𝗰𝗶𝗻𝗴 𝗮 𝘀𝘁𝗲𝗽 𝗯𝗮𝗰𝗸 from the immediate problem • 𝗣𝗿𝗼𝘃𝗶𝗱𝗶𝗻𝗴 𝘀𝘆𝘀𝘁𝗲𝗺 𝗰𝗼𝗻𝘁𝗲𝘅𝘁 that isn't visible from inside the code • 𝗞𝗻𝗼𝘄𝗶𝗻𝗴 𝘄𝗵𝗲𝗻 𝘁𝗼 𝘃𝗮𝗹𝗶𝗱𝗮𝘁𝗲 before declaring victory These are mentorship skills. Built by guiding people, sharpened by learning from them. Turns out, that's exactly what AI needs from us. What's the skill that's transferred best for you?