Skill Creator by hoangnguyen0403/agent-skills-standard
npx skills add https://github.com/hoangnguyen0403/agent-skills-standard --skill 'Skill Creator'编写高密度智能体技能的严格指南。通过渐进式披露和策略性内容组织,最大化信息密度,同时最小化令牌消耗。
每个单词都消耗令牌。设计技能以实现最大信息/令牌比:
references/ 文件夹Level 1: 元数据 (100 词) → 始终加载
Level 2: SKILL.md 主体 (100 行) → 触发时加载
Level 3: 引用/脚本/资源 → 按需加载
skills/
└── {category}/ # 例如:"flutter" (小写)
└── {skill-name}/ # 例如:"bloc-state-management" (短横线命名法)
├── SKILL.md # 核心逻辑 (高密度,<100 行)
├── scripts/ # 可执行代码 (确定性任务)
│ └── automation.py
├── references/ # 重量级示例 (延迟加载)
│ ├── patterns.md
│ └── examples.md
└── assets/ # 输出模板 (永不加载)
└── template.json
广告位招租
在这里展示您的产品或服务
触达数万 AI 开发者,精准高效
命令式压缩:以动词开头。不使用"请/你应该"。
令牌经济:最大化信息/令牌比。
渐进式披露:必要信息优先,细节按需提供。
上下文感知设计:不同智能体有不同的限制。
SKILL.md 中必需的部分:
前置元数据(必需):用于触发的元数据(最多 100 词)
name: 技能名称
description: 功能描述 + 使用时机 (触发激活)
metadata:
labels: [标签1, 标签2]
triggers:
files: ['**/*.ext']
keywords: [术语1, 术语2]
---
2. 优先级:P0(关键)、P1(标准)或 P2(可选)
结构:预期文件布局的 ASCII 树形图
指南:"执行此操作"的项目符号列表(命令式)
反模式:"不要执行此操作"的项目符号列表
参考链接:指向 references/ 文件的链接(延迟加载)
| 元素 | 限制 | 超出时操作 |
|---|---|---|
| SKILL.md 总计 | 100 行 | 提取到 references/ |
| 内联代码块 | 10 行 | 移至 references/ |
| 反模式项 | 15 词 | 压缩为命令式 |
| 优先级后的描述 | 0 行 | 移除 (使用前置元数据) |
| 表格 | 8 行 | 提取到 references/ |
| 解释性部分 | 10 行 | 提取到 references/ |
使用时机:确定性的、重复的任务 优点:永不加载到上下文中,直接执行 示例:代码生成器、格式化工具、验证器
使用时机:详细示例、API 文档、复杂模式 优点:仅在需要时加载,保持 SKILL.md 精简 示例:实现模式、错误处理指南
使用时机:样板文件、图像、配置 优点:永不加载,按需复制到输出 示例:项目模板、配置文件、图标
最终确定前,验证:
## 优先级 后重复前置元数据中的 description**粗体**:\更多粗体 `` - 造成视觉噪音格式:**禁止 X**:执行 Y[, 而非 Z]。[可选上下文,总计最多 15 词]
示例:
❌ 冗长 (24 词):
- **禁止手动发出**:`**避免 .then()**:不要在 Future.then 内部调用 emit();始终使用 await 或 emit.forEach。`
✅ 压缩 (11 词):
- **禁止 .then()**:使用 `await` 或 `emit.forEach()` 来发出状态。
❌ 冗长 (18 词):
- **禁止 UI 逻辑**:`**构建器中的逻辑**:不要在 BlocBuilder 内部执行计算或数据格式化。`
✅ 压缩 (9 词):
- **禁止在构建器中执行逻辑**:在 BLoC 中执行计算,而非 UI 中。
在以下情况提取到 references/:
使用下面的增强模板生成新技能:references/TEMPLATE.md
获取全面的生命周期指导:references/lifecycle.md
获取资源组织模式:references/resource-organization.md
每周安装次数
–
代码仓库
GitHub 星标数
355
首次出现
–
安全审计
Strict guidelines for authoring High-Density Agent Skills. Maximize information density while minimizing token consumption through progressive disclosure and strategic content organization.
Every word costs tokens. Design skills for maximum information/token ratio:
references/ folderLevel 1: Metadata (100 words) → Always loaded
Level 2: SKILL.md Body (100 lines) → When triggered
Level 3: References/Scripts/Assets → As needed
skills/
└── {category}/ # e.g., "flutter" (lowercase)
└── {skill-name}/ # e.g., "bloc-state-management" (kebab-case)
├── SKILL.md # Core Logic (High Density, <100 lines)
├── scripts/ # Executable code (Deterministic tasks)
│ └── automation.py
├── references/ # Heavy Examples (Lazy loaded)
│ ├── patterns.md
│ └── examples.md
└── assets/ # Output templates (Never loaded)
└── template.json
Imperative Compression : Start with verbs. No "Please/You should".
Token Economy : Maximize info/token ratio.
Progressive Disclosure : Essential info first, details on-demand.
Context-Aware Design : Different agents have different limits.
Required sections in SKILL.md:
Frontmatter (Mandatory) : Metadata for triggering (100 words max)
name: Skill Name
description: What it does + when to use it (triggers activation)
metadata:
labels: [tag1, tag2]
triggers:
files: ['**/*.ext']
keywords: [term1, term2]
---
2. Priority : P0 (Critical), P1 (Standard), or P2 (Optional)
Structure : ASCII tree of expected file layout
Guidelines : Bullet points of "Do this" (imperative)
Anti-Patterns : Bullet points of "Don't do this"
Reference Links : Links to references/ files (lazy loading)
| Element | Limit | Action if Exceeded |
|---|---|---|
| SKILL.md total | 100 lines | Extract to references/ |
| Inline code block | 10 lines | Move to references/ |
| Anti-pattern item | 15 words | Compress to imperative |
| Description after Priority | 0 lines | Remove (use frontmatter) |
| Tables | 8 rows | Extract to references/ |
| Explanatory sections | 10 lines | Extract to references/ |
When to use : Deterministic, repeated tasks Benefits : Never loaded into context, executed directly Examples : Code generators, formatters, validators
When to use : Detailed examples, API docs, complex patterns Benefits : Loaded only when needed, keeps SKILL.md lean Examples : Implementation patterns, error handling guides
When to use : Boilerplate files, images, configs Benefits : Never loaded, copied to output as-needed Examples : Project templates, config files, icons
Before finalizing, verify:
description after ## Priority**Bold**: \More Bold `` - causes visual noiseFormat: **No X**: Do Y[, not Z]. [Optional context, max 15 words total]
Examples :
❌ Verbose (24 words):
- **No Manual Emit**: `**Avoid .then()**: Do not call emit() inside Future.then; always use await or emit.forEach.`
✅ Compressed (11 words):
- **No .then()**: Use `await` or `emit.forEach()` to emit states.
❌ Verbose (18 words):
- **No UI Logic**: `**Logic in Builder**: Do not perform calculations or data formatting inside BlocBuilder.`
✅ Compressed (9 words):
- **No Logic in Builder**: Perform calculations in BLoC, not UI.
Extract to references/ when:
Use the enhanced template below to generate new skills: references/TEMPLATE.md
For comprehensive lifecycle guidance: references/lifecycle.md
For resource organization patterns: references/resource-organization.md
Weekly Installs
–
Repository
GitHub Stars
355
First Seen
–
Security Audits
agent-browser 浏览器自动化工具 - Vercel Labs 命令行网页操作与测试
147,400 周安装