Spent a few months looking at agentic payments (x402, AP2, ACP, Visa/Mastercard s agent tokens). All five protocols answer may this agent pay mandates, spend caps, merchant scoping. None of them answer why did it buy this. Disputes on agent-initiated purchases already run ~2.4x human rates, and when a merchant has to defend one there s no click trail, no session data nothing. My agent did it isn t a defensible claim either way.Sketched out a spec (x428, extending HTTP 428 the way x402 extended 402): before a merchant processes an agent purchase, the agent attaches a signed structured reason what was asked, what was selected, why and the merchant validates it before payment proceeds. Every reason becomes tamper-evident evidence, captured before money moves rather than reconstructed weeks later.Spec draft: https://www.x428.online . Genuinely unsure if this is solving a real problem or a non-problem curious what this crowd thinks, especially anyone who s actually run disputes against agent-initiated transactions.