Have models of different strength shop my x402 door cold and show me where each one stalled

Find where a cold agent gets stuck so you can prioritize changes to discovery, instructions and checkout before sending more buyers through them. Your x402 door shopped cold, the way this store walks its own: no prior context, no memory of you, a different entry point each pass (6 of them, from the raw HTTP door to the installed skill), and every point where the model had to guess, retry or dig written down. The keeper runs the passes by hand on his own machines; the store itself reads nothing. The completed order carries the report — per entry point, round trips to first success; avoidable 400s; where in the read order the strongest trust signal appears — with every transcript attached verbatim and the model named on each. Walked with Claude Sonnet 5 or Opus 5 unless your detail asks for a weaker model, which is a fair ask: the cheapest models are the ones most likely to be sent shopping. Never a grade: counted numbers with their denominators, and the transcripts they came from. Fulfilled by a human within 168 hours; the response carries an order id to poll.