Claude外部上下文技能：智能分解查询，并行搜索技术文档与参考资料

external-context by yeachan-heo/oh-my-claudecode

165 周安装量

24,900 GitHub Stars

GitHub

安装命令

npx skills add https://github.com/yeachan-heo/oh-my-claudecode --skill external-context

AI/机器学习开发搜索

🇨🇳中文介绍

外部上下文技能

获取查询的外部文档、参考资料和上下文。将查询分解为 2-5 个方面，并并行启动文档专家 Claude 智能体。

使用方法

/oh-my-claudecode:external-context <主题或问题>

示例

/oh-my-claudecode:external-context Node.js 中 JWT 令牌轮换的最佳实践是什么？
/oh-my-claudecode:external-context 比较 Prisma 与 Drizzle ORM 在 PostgreSQL 上的表现
/oh-my-claudecode:external-context 最新的 React 服务器组件模式和约定

协议

步骤 1：方面分解

给定一个查询，将其分解为 2-5 个独立的搜索方面：

## 搜索分解

**查询:** <原始查询>

### 方面 1: <方面名称>
- **搜索重点:** 要搜索的内容
- **来源:** 官方文档、GitHub、博客等

### 方面 2: <方面名称>
...

步骤 2：并行智能体调用

通过 Task 工具并行启动独立的方面：

Task(subagent_type="oh-my-claudecode:document-specialist", model="sonnet", prompt="搜索: <方面 1 描述>。使用 WebSearch 和 WebFetch 查找官方文档和示例。使用 URL 引用所有来源。")

Task(subagent_type="oh-my-claudecode:document-specialist", model="sonnet", prompt="搜索: <方面 2 描述>。使用 WebSearch 和 WebFetch 查找官方文档和示例。使用 URL 引用所有来源。")

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

联系我们

🇺🇸English

External Context Skill

Fetch external documentation, references, and context for a query. Decomposes into 2-5 facets and spawns parallel document-specialist Claude agents.

Usage

/oh-my-claudecode:external-context <topic or question>

Examples

/oh-my-claudecode:external-context What are the best practices for JWT token rotation in Node.js?
/oh-my-claudecode:external-context Compare Prisma vs Drizzle ORM for PostgreSQL
/oh-my-claudecode:external-context Latest React Server Components patterns and conventions

Protocol

Step 1: Facet Decomposition

Given a query, decompose into 2-5 independent search facets:

## Search Decomposition

**Query:** <original query>

### Facet 1: <facet-name>
- **Search focus:** What to search for
- **Sources:** Official docs, GitHub, blogs, etc.

### Facet 2: <facet-name>
...

Step 2: Parallel Agent Invocation

Fire independent facets in parallel via Task tool:

Task(subagent_type="oh-my-claudecode:document-specialist", model="sonnet", prompt="Search for: <facet 1 description>. Use WebSearch and WebFetch to find official documentation and examples. Cite all sources with URLs.")

Task(subagent_type="oh-my-claudecode:document-specialist", model="sonnet", prompt="Search for: <facet 2 description>. Use WebSearch and WebFetch to find official documentation and examples. Cite all sources with URLs.")

Maximum 5 parallel document-specialist agents.

Step 3: Synthesis Output Format

Present synthesized results in this format:

## External Context: <query>

### Key Findings
1. **<finding>** - Source: [title](url)
2. **<finding>** - Source: [title](url)

### Detailed Results

#### Facet 1: <name>
<aggregated findings with citations>

#### Facet 2: <name>
<aggregated findings with citations>

### Sources
- [Source 1](url)
- [Source 2](url)

Configuration

Maximum 5 parallel document-specialist agents
No magic keyword trigger - explicit invocation only

Weekly Installs

Repository

yeachan-heo/oh-…audecode

GitHub Stars

11.3K

First Seen

Feb 14, 2026

Security Audits

Gen Agent Trust HubPass SocketPass SnykWarn

Installed on

opencode94

codex93

gemini-cli92

claude-code91

cursor91

kimi-cli89