Name: token-optimization
Author: YonganZhang

Imported from YonganZhang/code-token-optiization (examples/skills/token-optimization/SKILL.md). Install upstream with npx skills add YonganZhang/code-token-optiization --skill token-optimization. Copyright stays with the author.

Token 优化体系

通过 Rule/Skill/MCP 分层管理实现 token 节省的完整方法论。本 skill 自身即采用多文件渐进式披露结构。

核心原则

固定成本最小化 — Rule/CLAUDE.md 每轮全文加载，必须精简；Skill 按需加载，存详情
渐进式披露 — 信息分层：Rule(触发索引) → Skill SKILL.md(目录) → references/(详情)
元信息优于内容 — 存"哪里有什么"的索引，不存会变的实现细节

体系架构（四层）

层级	载体	加载时机	内容
L0	Rule（≤10行）	每轮	触发条件 + 指向 Skill 的索引
L1	Skill SKILL.md（≤50行）	触发时	目录 + 各子文件一句话摘要
L2	Skill references/*.md	AI 按需 Read	详细规则、模板、示例
L3	源码	AI 按需 Read	实现细节（不存储，直接读源文件）

参考文档目录

文件	内容	何时读
`references/cost-model.md`	上下文成本模型、MCP Tool Search、Rule 精简	理解原理时
`references/patterns.md`	Rule+Skill 联动、多文件披露、项目文档拆分、索引自维护	创建 Rule/Skill 时
`references/setup.md`	从零搭建步骤、当前文件清单、新机器复用	首次搭建或迁移时

快速入口

首次搭建 → 读 references/setup.md
理解成本模型 → 读 references/cost-model.md
创建新 Rule/Skill/项目文档 → 读 references/patterns.md

# after Install: the listing is in your workspace's routing pool - a plan picks it for its slot curl -s -X POST https://api.opensmartroute.ai/api/v1/route -H 'Authorization: Bearer $OSR_API_KEY' -H 'Content-Type: application/json' -d '{"text": "...", "plan": true}'

Install token-optimization and it becomes one more candidate the router can pick - when it fits.

A listing is a routing target with a manifest: what it does, which domains it covers, what it costs and who publishes it. Once installed it sits beside your own models and tools, is scored like any other candidate for each request, and shows up in the trace when it wins. Ratings come from workspaces that installed it, one per account.

{ "ocm": "1", "id": "yonganzhang-code-token-optiization-token-optimization", "kind": "skill", "name": "token-optimization", "description": "Claude Code Token 优化体系 — Rule+Skill联动、渐进式披露、MCP按需加载，可从零复现的完整方法论", "publisher": "YonganZhang", "version": "1.0.0", "capabilities": { "domains": [ "coding" ], "tags": [ "skill-md", "github" ], "languages": [ "en" ] }, "quality_prior": 0.6, "examples": [ "Claude Code Token 优化体系 — Rule+Skill联动、渐进式披露、MCP按需加载，可从零复现的完整方法论" ], "primary": false, "metadata": { "source": { "provider": "github", "repository": "https://github.com/YonganZhang/code-token-optiization", "path": "examples/skills/token-optimization/SKILL.md", "ref": "0d69ef7745d0ba67d492c734e079b8d5ab662112", "url": "https://github.com/YonganZhang/code-token-optiization/blob/0d69ef7745d0ba67d492c734e079b8d5ab662112/examples/skills/token-optimization/SKILL.md", "key": "YonganZhang/code-token-optiization/examples/skills/token-optimization/SKILL.md" } }, "instructions": "# Token 优化体系\n\n通过 Rule/Skill/MCP 分层管理实现 token 节省的完整方法论。本 skill 自身即采用多文件渐进式披露结构。\n\n## 核心原则\n\n1. **固定成本最小化** — Rule/CLAUDE.md 每轮全文加载，必须精简；Skill 按需加载，存详情\n2. **渐进式披露** — 信息分层：Rule(触发索引) → Skill SKILL.md(目录) → references/(详情)\n3. **元信息优于内容** — 存\"哪里有什么\"的索引，不存会变的实现细节\n\n## 体系架构（四层）\n\n| 层级 | 载体 | 加载时机 | 内容 |\n|------|------|---------|------|\n| L0 | Rule（≤10行） | 每轮 | 触发条件 + 指向 Skill 的索引 |\n| L1 | Skill SKILL.md（≤50行） | 触发时 | 目录 + 各子文件一句话摘要 |\n| L2 | Skill references/*.md | AI 按需 Read | 详细规则、模板、示例 |\n| L3 | 源码 | AI 按需 Read | 实现细节（不存储，直接读源文件） |\n\n## 参考文档目录\n\n| 文件 | 内容 | 何时读 |\n|------|------|--------|\n| `references/co", "cost": { "context_tokens": 232 } }

token-optimization

About

Token 优化体系

核心原则

体系架构（四层）

参考文档目录

快速入口

Use it

Install token-optimization and it becomes one more candidate the router can pick - when it fits.

Manifest

Reviews