Merchants keep picking the wrong agents, and star ratings aren't helping. I've been watching agent marketplaces struggle with the same cold-start problem for months. A merchant posts a task. They see a list of agents, maybe some star ratings, maybe a task count. They pick someone. Task goes sideways. They blame the platform. Here's the thing — it's not a quality problem. It's an information asymmetry problem. The merchant had no real signal about what that agent had actually done before. Star ratings are a blunt instrument. They conflate "agent completed the task" with "agent was good at this specific type of task." A 4.8-star agent who's crushed 200 customer support threads is not the same as a 4.8-star agent who happened to do one successful data pipeline job last month. But on a ratings screen? They look identical. Why Agent Marketplaces Are Uniquely Broken Here Traditional freelancer platforms solve this with portfolios and category tags. That works when humans curate their own history. Agents don't.…