跳到主要内容
知仓学习社ZHICANG

router-first-architecture

Router-First Architecture

不碰外部(只输出文字)无严重或高危命中parcadei/Continuous-Claude-v3

它会碰到什么

扫了多少1 个文本文件,2 KB
它会碰到什么不碰外部(只输出文字)
命中总数0 处
命中统计严重 0 · 高 0 · 中 0 · 低 0

这一栏是扫描器报的事实,不是结论。命中多不等于有毒(安全工具、规则库、示例脚本本来就会包含危险写法),命中少也不等于干净。它和你手上的凭据、文件、网络有什么关系,需要你自己看。

技能内容

Router-First Architecture

Route through domain routers before using individual tools. Routers abstract tool selection.

Pattern

Domain routers (like math-router) provide deterministic mapping from user intent to exact CLI commands. Always use the router first; only bypass for edge cases.

DO

  • Call math-router route "<intent>" before any math operation
  • Let domain skills co-activate with their router (via coActivate in skill-rules.json)
  • Trust the router's confidence score; only fall back if command: null
  • Keep trigger keywords/patterns in skill-rules.json broader than routing patterns

DON'T

  • Call individual scripts directly when a router exists
  • Duplicate routing logic in individual skills
  • Let domain skills bypass their router

Co-Activation Pattern

Domain skills should co-activate with their router:

{
  "math/abstract-algebra/groups": {
    "coActivate": ["math-router"],
    "coActivateMode": "always"
  }
}

This ensures the router is always available when domain knowledge is activated.

Two-Layer Architecture

  1. Skill-rules trigger layer: Nudges Claude to use the router (keywords, intent patterns)
  2. Router routing layer: Deterministic mapping to scripts via regex patterns

Keep the trigger layer broader than routing - the router should handle "not found" gracefully.

Source Sessions

  • 2bbc8d6e: "Trigger layer was narrower than routing layer" - expanded triggers
  • This session: Wired 8 domain math skills to co-activate with math-router

想直接用这个技能?

本站把开放许可(MIT / Apache 等)的技能按仓库打包整理到网盘,点一下转存到你自己的网盘,不用一个个从 GitHub 拉。许可未声明的技能只给原始仓库链接,不打包。

同名技能的其他版本

有 2 个不同仓库或目录里都有叫 router-first-architecture 的技能。它们内容并不相同,别混用: