Claude Code 正在经历一次架构层面的自我重构——其核心功能模块 Function Hooks 已正式从内部实验阶段毕业，以 **Mods** 的全新身份对外发布。这标志着 Claude Code 从“单体智能体”向“可插拔模块化平台”的演进迈出了关键一步。 ## 从 Function Hooks 到 Mods：一次身份升级 Function Hooks 原本是 Claude Code 内部用于拦截、增强和扩展工具调用行为的底层机制。它允许开发者在 Claude 执行特定函数（如读写文件、运行命令、调用 API）前后注入自定义逻辑，从而实现权限控制、日志审计、输入输出改写等能力。 此次更名为 **Mods**，不仅仅是换了个名字，而是将其定位从“内部钩子”提升为“一等公民扩展单元”。Mods 现在具备： - **独立生命周期**：可单独安装、启用、禁用、卸载，不再依附于特定工具链。 - **标准化接口**：统一的 manifest 描述文件，声明触发条件、执行时机、依赖关系。 - **组合能力**：多个 Mods 可串联或并联，形成处理管道。 - **沙箱隔离**：每个

动察 Beating AI News Flash: Anthropic has officially launched Claude Code Mods. This is the previously previewed Function Hooks, and the official product name is now Mods. Under the hood, it still uses Function Hooks; a Mod is essentially a plugin that can directly modify Claude Code.

Mods can change Claude Code's behavior and interface. They can rewrite content before prompts are sent to the model, intercept, modify, or retry tool calls, handle permission requests, and even replace parts of the interface. Users can also directly ask Claude to write a Mod, install it, and load it immediately.

Anthropic has also begun splitting Claude Code's built-in features into Mods. /diff has already been migrated first, and users can disable it or replace it with their own version. The official plan is to continue migrating more built-in features so that Claude Code ultimately retains only a smaller core. Mods are still installed and shared through the existing Plugin system, and can be used on both CLI and desktop.

However, Mods do not have sandbox isolation and have the same local permissions as Claude Code itself. Anthropic recommends installing only Mods from trusted sources.