research

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Session: ${CLAUDE_SESSION_ID}

Research Topic: $ARGUMENTS

会话: ${CLAUDE_SESSION_ID}

研究主题: $ARGUMENTS

Research Skill

调研技能

Research Workflows

调研工作流

Standard Query

标准查询流程

1. THINK → Sequential Thinking decomposition
2. RESOLVE → Context7 resolve-library-id
3. DOCUMENT → Context7 query-docs (5000-10000 tokens)
4. SUPPLEMENT → Exa code context search
5. SYNTHESIZE → Structured answer with sources

1. THINK → Sequential Thinking 分解问题
2. RESOLVE → 调用Context7的resolve-library-id接口
3. DOCUMENT → 调用Context7的query-docs接口（返回5000-10000 tokens）
4. SUPPLEMENT → 使用Exa进行代码上下文搜索
5. SYNTHESIZE → 生成带来源标注的结构化答案

Complex Investigation

复杂调查流程

1. DEEP THINK → Multi-hypothesis Sequential Thinking
2. DEEP RESEARCH → Exa deep researcher (45s-2min)
3. MONITOR → Check status until completed
4. VALIDATE → Cross-check Context7 official sources
5. REPORT → Comprehensive solution

1. DEEP THINK → 基于多假设的Sequential Thinking
2. DEEP RESEARCH → 启动Exa深度调研（耗时45秒-2分钟）
3. MONITOR → 持续检查任务状态直至完成
4. VALIDATE → 通过Context7官方来源交叉验证信息
5. REPORT → 输出完整的解决方案

Technology Trends

技术趋势调研流程

1. WEB SCAN → Exa search latest developments
2. CODE PATTERNS → Exa code context for practices
3. ECOSYSTEM → Company research for key players
4. ANALYSIS → Sequential Thinking for implications
5. RECOMMENDATIONS → Actionable insights

1. WEB SCAN → 使用Exa搜索最新技术进展
2. CODE PATTERNS → 通过Exa搜索代码实践案例
3. ECOSYSTEM → 调研行业关键参与企业
4. ANALYSIS → 用Sequential Thinking分析技术影响
5. RECOMMENDATIONS → 输出可落地的洞察建议

Context7 Usage

Context7 使用指南

typescript

// Step 1: Resolve library ID
mcp__context7__resolve-library-id({
  libraryName: "next.js",
  query: "App Router server actions"
})

// Step 2: Query docs
mcp__context7__query-docs({
  libraryId: "/vercel/next.js",
  query: "server actions authentication"
})

Best Practices:

Always
```
resolve-library-id
```
BEFORE
```
query-docs
```
Specify
```
topic
```
parameter to focus retrieval
Start with 5000 tokens, increase to 10000 if needed
Handle variations: "nextjs" vs "/vercel/next.js"

typescript

// Step 1: Resolve library ID
mcp__context7__resolve-library-id({
  libraryName: "next.js",
  query: "App Router server actions"
})

// Step 2: Query docs
mcp__context7__query-docs({
  libraryId: "/vercel/next.js",
  query: "server actions authentication"
})

最佳实践:

务必先调用
```
resolve-library-id
```
，再调用
```
query-docs
```
指定
```
topic
```
参数以聚焦检索范围
初始请求设置为5000 tokens，若信息不足可增加至10000 tokens
处理库名变体：比如“nextjs”与“/vercel/next.js”的差异

Exa Search Types

Exa 搜索类型

Type	Use Case	Time	numResults
`fast`	Quick lookups	<5s	3-5
`auto`	Balanced	5-15s	5-8
`deep`	Comprehensive	15-45s	8+

typescript

// Code context search
mcp__exa__get_code_context_exa({
  query: "Next.js 16 server actions authentication",
  tokensNum: 5000
})

// Web search
mcp__exa__web_search_exa({
  query: "React 2025 best practices",
  type: "auto",
  numResults: 5
})

类型	适用场景	耗时	结果数量
`fast`	快速查询	<5秒	3-5条
`auto`	平衡型查询	5-15秒	5-8条
`deep`	全面调研	15-45秒	8条以上

typescript

// 代码上下文搜索
mcp__exa__get_code_context_exa({
  query: "Next.js 16 server actions authentication",
  tokensNum: 5000
})

// 网页搜索
mcp__exa__web_search_exa({
  query: "React 2025 best practices",
  type: "auto",
  numResults: 5
})

Exa Deep Research

Exa深度调研

Reserve for investigations requiring >30min manual effort.

typescript

// Start research
const { taskId } = await mcp__exa__deep_researcher_start({
  instructions: "Compare authentication solutions for Node.js",
  model: "exa-research-pro" // or "exa-research" for faster
})

// Poll until complete
mcp__exa__deep_researcher_check({ taskId })

Models:

```
exa-research
```
: Standard depth (15-45s)
```
exa-research-pro
```
: Complex topics (45s-2min)

适用场景: 手动完成需耗时30分钟以上的调查任务。

typescript

// 启动调研任务
const { taskId } = await mcp__exa__deep_researcher_start({
  instructions: "Compare authentication solutions for Node.js",
  model: "exa-research-pro" // 若需更快速度可使用"exa-research"
})

// 轮询任务状态直至完成
mcp__exa__deep_researcher_check({ taskId })

可用模型:

```
exa-research
```
: 标准深度（15-45秒）
```
exa-research-pro
```
: 复杂主题调研（45秒-2分钟）

Sequential Thinking

Sequential Thinking 分步思考

typescript

mcp__sequential-thinking__sequentialthinking({
  thought: "Analyzing authentication approaches",
  thoughtNumber: 1,
  totalThoughts: 5,
  nextThoughtNeeded: true,
  // Optional for revisions:
  isRevision: false,
  revisesThought: null,
  branchId: null,
  branchFromThought: null,
  needsMoreThoughts: false
})

Best Practices:

Start with realistic
```
totalThoughts
```
, adjust dynamically
Use
```
isRevision: true
```
to reconsider hypotheses
Create branches (
```
branchId
```
) for alternatives
Set
```
needsMoreThoughts: true
```
if incomplete

typescript

mcp__sequential-thinking__sequentialthinking({
  thought: "Analyzing authentication approaches",
  thoughtNumber: 1,
  totalThoughts: 5,
  nextThoughtNeeded: true,
  // 修订时可选参数:
  isRevision: false,
  revisesThought: null,
  branchId: null,
  branchFromThought: null,
  needsMoreThoughts: false
})

最佳实践:

初始设置合理的
```
totalThoughts
```
数值，可根据实际情况动态调整
若需重新考虑假设，设置
```
isRevision: true
```
可创建分支（
```
branchId
```
）以探索替代方案
若信息不完整，设置
```
needsMoreThoughts: true
```

Multi-Source Synthesis

多源信息合成

Parallelization:

Run
```
resolve-library-id
```
+
```
web_search_exa
```
simultaneously
Launch multiple Exa searches concurrently
Execute Context7 docs + Exa code search in parallel

Source Priority:

Official documentation (Context7)
Recent tutorials (Exa, <6 months)
Older content (with version verification)

并行化策略:

同时调用
```
resolve-library-id
```
和
```
web_search_exa
```
并发启动多个Exa搜索任务
并行执行Context7文档查询与Exa代码上下文搜索

信息源优先级:

官方文档（Context7）
近期教程（Exa，发布时间<6个月）
旧版内容（需验证版本兼容性）

Response Format

响应格式

markdown

undefined

markdown

undefined

🔍 Research: [Topic]

🔍 调研结果: [主题]

Methodology

调研方法

Sequential Thinking: [N thoughts, M revisions]
Context7: [Library@version consulted]
Exa: [Search types performed]

Sequential Thinking: [共N个思考步骤，M次修订]
Context7: [查阅的库@版本信息]
Exa: [执行的搜索类型]

Key Findings

关键发现

[Finding 1] (Source: [URL])
- Technical details
- Code examples
[Finding 2] (Source: [URL])

[发现1] (来源: [URL])
- 技术细节
- 代码示例
[发现2] (来源: [URL])

Recommendations

建议

[Action 1]: [Why + How]
[Action 2]: [Why + How]

[行动1]: [原因 + 实施方式]
[行动2]: [原因 + 实施方式]

Sources

信息来源

Context7: [Exact library IDs]
Exa: [X results analyzed]
Deep Research: [Task ID if used]

undefined

Context7: [具体库ID]
Exa: [分析了X条结果]
深度调研: [若使用则填写任务ID]

undefined

Error Handling

错误处理

Context7 Failures:

Verify library name spelling
Try different formats ("/org/project" vs "project-name")
Fallback to Exa code context

Exa Timeouts:

Reduce
```
numResults
```
Simplify query
Switch
```
type: "deep"
```
→
```
type: "fast"
```

Sequential Thinking Blocks:

Revise with
```
isRevision: true
```
Increase
```
totalThoughts
```
Create new branch

Context7 调用失败:

验证库名拼写是否正确
尝试不同的格式（比如“/org/project”与“project-name”）
fallback到Exa代码上下文搜索

Exa 请求超时:

减少
```
numResults
```
数量
简化查询语句
将搜索类型从
```
type: "deep"
```
切换为
```
type: "fast"
```

Sequential Thinking 陷入停滞:

设置
```
isRevision: true
```
重新思考
增加
```
totalThoughts
```
数值
创建新的思考分支

Forbidden Behaviors

禁止行为

❌ Guess library IDs without
```
resolve-library-id
```
❌ Start deep researcher without checking completion
❌ Mix opinions with facts without distinction
❌ Provide code without version verification
❌ Ignore WebFetch redirects
❌ Recommend without citing sources
❌ Skip Sequential Thinking for multi-step problems

❌ 未调用
```
resolve-library-id
```
就直接猜测库ID
❌ 未检查任务完成状态就重复启动深度调研
❌ 未加区分地将主观观点与客观事实混合输出
❌ 提供未验证版本兼容性的代码
❌ 忽略WebFetch重定向
❌ 推荐内容时不标注信息来源
❌ 处理多步骤问题时跳过Sequential Thinking环节