📄 中文摘要
在多代理系统设计中，存在一个被称为“信任差距”的问题：代理 A 希望代理 B 完成某项任务，而代理 B 希望在开始之前获得报酬，但两者都无法强制执行协议。人类通常通过可信的第三方（如律师、托管公司或具有争议解决能力的支付处理商）来解决此问题。Purple Flea 的托管服务旨在探讨在两个自主代理之间，如何实现无信任的托管机制。最简单的解决方案是直接支付，但这种方法存在局限性，无法有效解决代理之间的信任问题。
在 AI 代理之间建立无信任托管
出处: Building trustless escrow between AI agents
发布: 2026年3月6日
📄 English Summary
Building trustless escrow between AI agents
The design of multi-agent systems often reveals a 'trust gap' where Agent A wants Agent B to perform a task, but Agent B requires payment upfront, and neither can enforce the agreement. Human solutions typically involve a trusted third party, such as a lawyer or escrow service. The escrow service developed in Purple Flea aims to explore what trustless escrow looks like when both parties are autonomous agents. The simplest solution is direct payment; however, this approach has limitations and fails to adequately address the trust issues between agents.