Exgentic/agent-cards

Agent Cards是一个标准化的格式，用于记录AI代理系统——包括它们是什么、如何工作以及性能如何。它描述了完整的代理系统，涵盖规划策略、工具使用、内存、错误恢复以及内部使用的模型。该格式旨在使不同代理系统在成本、失败模式和架构方面的差异可见，从而促进对代理评估的深入理解。它包括机器可读的前言（如身份信息、框架、模型和标签）和人类可读的部分（如代理详情、架构、模型、支持环境、评估结果、局限性和运行指南）。Agent Cards is a standardized format for documenting AI agent systems, covering their definitions, operational mechanisms, and performance characteristics. It describes complete agent systems, including planning strategies, tool utilization, memory management, error recovery, and internally deployed models. This format is designed to make the disparities across different agent systems visible in terms of cost, failure modes, and architecture, thereby facilitating in-depth insights into agent evaluation. It consists of machine-readable preambles (such as identity information, frameworks, models, and tags) and human-readable sections (including agent details, architecture, models, supported environments, evaluation results, limitations, and operational guidelines).