代理型人工智能基金会（AAIF）已被委托管理另外四个AI项目，以期扩大用于制作AI代理的厂商中立工具集。AAIF于2025年12月在Linux基金会的支持下成立，是一个厂商中立的组织，旨在造福那些更热衷于“水涨船高”而非“翻江倒海”的AI公司。它成立时监管着三个项目：模型上下文协议（MCP）、goose和AGENTS.md。
The Agentic AI Foundation (AAIF) has been entrusted with stewardship of four additional AI projects in the hope of broadening the set of vendor-neutral tools for making AI agents. The AAIF was founded in December 2025 under the auspices of the Linux Foundation as a vendor-neutral organization for the benefit of AI companies more interested in riding a tide that lifts all boats than making waves. It launched with three projects under its supervision: Model Context Protocol (MCP), goose, and AGENTS.md.
随着四个新项目的加入，现在有10个项目在AAIF的庇护下。捐赠组织——阿里巴巴、Anthropic、Hugging Face和OpenAI——已分别移交了OpenSandbox、Agent Skills、OpenEnv和Open Responses的控制权。
With the four new adoptees, there are now 10 projects sheltering under the AAIF's wings. The donor organizations – Alibaba, Anthropic, Hugging Face, and OpenAI – have handed over control of OpenSandbox, Agent Skills, OpenEnv, and Open Responses respectively.
这些项目中的每一个都提供了一些功能，AI代理的开发者（通常结合了模型、工具和任务循环）可能会觉得这些功能很有用。而且，拥有一个或多或少达成共识的实现蓝图往往会使部署变得更加容易和可预测，至少在你的AI代理意外洗劫某个第三方组织之前是这样的。OpenSandbox提供了一种方法，通过使用各种客户端SDK、工具、协议、生命周期钩子、容器运行时以及各种沙盒内执行组件来限制或沙盒化AI代理的功能。
Each of these projects provides functions that developers of AI agents – which typically combine models, tools, and a task loop – may find helpful. And having a more or less agreed-upon blueprint for implementation tends to make deployment easier and more predictable, at least until your AI agent unexpectedly ransacks some third-party organization. OpenSandbox provides a way to constrain or sandbox the capabilities of AI agents using various client SDKs, tools, protocols, lifecycle hooks, container runtimes, and assorted in-sandbox execution components.
Agent Skills概述了一种文本描述格式，可以更有效地引导AI代理完成指定的任务。它提供了有助于代理实现目标的上下文。
Agent Skills outlines a format for text descriptions that guide AI agents through specified tasks more effectively. It provides context that helps agents accomplish their goals.
AAIF的联合创始人Anthropic在一篇博客文章中表示：“我们致力于保持Agent Skills的开放性、社区驱动和厂商中立。”OpenEnv提供了一项容器化服务，用于创建强化学习环境，该环境可以毫不费力地构建、可以互操作，并且可以扩展以满足严苛的测试需求。Open Responses则是一个开放的API规范，用于构建可与不同模型提供商合作的LLM接口。它规范了常见的请求和响应格式、流式事件以及工具调用模式。
"We’re committed to keeping Agent Skills open, community-driven, and vendor-neutral," said Anthropic, an AAIF co-founder, in a blog post. OpenEnv offers a containerized service for creating reinforcement learning environments that can be built without much trouble, can interoperate, and can scale to meet demanding testing. And Open Responses is an open API specification for building LLM interfaces that can work with different model providers. It standardizes common request and response formats, streaming events, and tool invocation patterns.
AAIF对这些技术规范的培育吸引了大量公司的浓厚兴趣，这些公司坚信AI代理拥有未来。该组织在不到一年的时间里会员数量已超过300家，本季度新增了60家，其中包括作为黄金会员的蚂蚁集团、盖茨基金会、MongoDB、Nutanix、加拿大皇家银行、Vercel和Vultr。
The AAIF's nurturing of these technical specs has attracted considerable interest from companies convinced that there's a future in AI agents. The organization has surpassed 300 members in less than a year, with 60 coming on board this quarter, including Ant Group, Gates Foundation, MongoDB, Nutanix, Royal Bank of Canada, Vercel, and Vultr as Gold Members.
这种对AI代理的热情（或轻信）拥抱对某些行业来说可能节奏过快了。据Axios报道，美国银行已经开始担心，AI代理在未经银行授权的情况下与客户账户进行交互（尽管不一定未经客户授权）。®
This enthusiastic (or credulous) embrace of AI agents may be a bit too rapid for some industries. Axios is reporting that US banks have become concerned that AI agents are interacting with customer accounts without bank authorization (though not necessarily without customer authorization). ®