Hi Product Hunt, I am a designer. My team ships with coding agents every day, and the same thing kept happening: the agent hands back working code in minutes, then we spend the afternoon fixing the parts nobody told it about. Focus order. Error recovery. Touch target size. What the empty state actually says. The agent was not careless, it just had nothing authoritative to check against. So I built Rubric. It is a UX corpus served over MCP. 554 rules, each written as a rule rather than an opinion: what to do, why it matters, how bad it is if you skip it, and where it comes from. Your agent calls it before it writes the interface, so the guidance arrives while the decisions are still cheap to change. The part I care about most is that the same corpus runs the audit. Point it at a live URL and it checks the site against the exact rules it designed with, then returns findings ranked by severity, each with issue, impact, actions and references. No gap between the standard you build to and the standard you get measured against. Setup is a browser approval, no API key copying. The free tier gives you the full checklist with issue and severity data, 100 calls a day for free (for a short period of time) no card, credit card. Two things I would love feedback on: 1. Which rules feel too strict, or plain wrong, for the kind of product you build. 2. Whether the audit findings are actionable enough to hand straight back to an agent. I am here all day, ask me anything.