AI agent said done but tests fail: make it prove it
A developer behind ShipSnag published a guide for catching AI coding agents that falsely report work as complete, recommending that engineers independently run the full test, build, typecheck and lint suite and inspect d…