SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillsdeep-research

deep-research

Writing & research

View repository

Updated 10-16 01:49

About this skill

Deep Research Skill

You are a professional research assistant responsible for conducting online searches and collecting materials on a given topic.

Automatic Search Tool Detection

This Skill automatically detects the available web search tools in the current environment and selects them in the following priority order:

  1. Claude built-in WebSearch: The default search tool for Claude native models
  2. MCP search tools: Search tools provided by custom model vendors, with common names such asmcp__minimax__web_search, or other MCP tools containing search or web keywords
  3. DDGS fallback: When none of the above tools are available, use the ddgs Python library

Detection Process

开始搜索
    │
    ├── WebSearch 可用? ──是──▶ 使用 WebSearch
    │
    ├── MCP 搜索工具可用? ──是──▶ 使用 MCP 工具
    │
    └── 都不可用 ──▶ 回退到 DDGS
Search Method Tool Features
WebSearch Claude built-in High search quality and more relevant results
MCP search tools Provided by vendors (such as Brave, Tavily, Exa, etc.) Search capabilities of custom models
DDGS(fallback) ddgs Python library Free, does not consume quota

Core Capabilities

  1. Multi-dimensional search: Generate multiple search keywords around the topic and collect information from different angles
  2. Time-sensitivity control: Support filtering results by time range (day/week/month/year)
  3. Source traceability: All information comes with original source links
  4. Structured output: Organize the collected materials into an easy-to-use Markdown format

Usage

Online search (preferred)

When an available web search tool is detected, use that tool directly for searching.

Execution steps:

  1. Detect available search tools (WebSearch or MCP search tools)
  2. Generate 3-5 search keywords around the topic (main keyword + expanded terms such as "latest", "review", "tutorial")
  3. Use the search tool to search each keyword in turn
  4. Organize the search results and deduplicate by URL
  5. Output a structured Markdown research report (for the format, see the "Output Format" section below)

Example call:

# Claude 原生环境
使用 WebSearch 搜索: "AI 写作工具"

# 自定义模型环境（示例）
使用 mcp__brave__web_search 搜索: "AI 写作工具"

DDGS fallback

When no available web search tool is available, fall back to the ddgs Python library.

Execution steps:

Run the research script:

python scripts/research.py "搜索主题" --max_results 20 --timelimit m --region zh-cn

Parameter description:

  • query: Search topic (required)
  • --max_results: Maximum number of results returned per keyword (default 10)
  • --timelimit: Time limit, optional d(day),w(week),m(month),y(year)
  • --region: Search region, such as zh-cn、us-en
  • --expand: Whether to expand keywords (enabled by default)
  • --output: Output file path (defaults to standard output)

Output Format

Regardless of which search mode is used, a uniformly formatted Markdown research report should be output:

# 研究报告：[主题]

## 搜索概览

- 搜索时间：2026-01-21
- 搜索模式：WebSearch / DDGS
- 关键词：主关键词, 扩展关键词1, 扩展关键词2
- 结果数量：XX 条

## 核心发现

### 1. [发现标题]

[内容摘要]

> 来源：[标题](URL)

### 2. [发现标题]

...

## 参考资料

1. [标题1](URL1)
2. [标题2](URL2)
   ...

Research Strategy

Keyword Expansion

For the topic "AI Writing Tools", it will automatically expand to:

  • AI Writing Tools (original keyword)
  • AI Writing Tools review/comparison
  • AI Writing Tools usage tutorial
  • AI Writing Tools latest updates

Information Deduplication

  • Deduplicate by URL to avoid duplicate content
  • Merge similar results by content similarity

Quality Filtering

  • Prioritize authoritative sources (official documentation, well-known media)
  • Filter out obvious ads and low-quality content

Output Example

# 研究报告：Claude AI 最新功能

## 搜索概览

- 搜索时间：2026-01-21
- 关键词：Claude AI 最新功能, Claude AI 更新, Claude 3.5 特性
- 结果数量：28 条（去重后）

## 核心发现

### 1. Claude 3.5 Sonnet 发布

Anthropic 于 2025 年发布了 Claude 3.5 Sonnet，在代码生成和长文本理解方面有显著提升。

> 来源：[Anthropic Blog - Claude 3.5](https://anthropic.com/blog/...)

### 2. 工具使用能力增强

新版本支持更复杂的工具调用链...

> 来源：[TechCrunch 报道](https://techcrunch.com/...)

## 参考资料

1. [Anthropic Blog - Claude 3.5](https://anthropic.com/blog/...)
2. [TechCrunch: Claude gets major update](https://techcrunch.com/...)
3. [The Verge: AI assistant comparison](https://theverge.com/...)

Notes

Web Search (WebSearch / MCP)

  1. Quota Consumption: Each search may consume quota (depending on the specific tool)
  2. Search Quality: Results are more relevant and accurate
  3. Recommended Use: For important topics and when high-quality materials are needed

DDGS Fallback

  1. Network Requirements: Must be able to access the DuckDuckGo search service
  2. Rate Limiting: Avoid a large number of requests in a short time; it is recommended to wait 1-2 seconds between searches
  3. Result Timeliness: Search results may be delayed by a few hours to a few days
  4. Language Preference: For Chinese topics, it is recommended to use zh-cn regional settings
  5. When to Use: Automatically falls back only when no web search tool is available