SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillsmulti-search

multi-search

Writing & research

View repository

Updated 10-15 12:57

About this skill

Multi-Topic Deep Research Skill

Overview

An intelligent multi-topic deep research tool that automatically analyzes materials and generates systematic research documents. It supports any material input and, throughlaunching multiple independent research Agents in parallelconducts deep retrieval to form a concise research knowledge base.

Core Principles：

  • Only perform information retrieval, summarization, and professional expression transformation
  • Do not add new facts, do not fabricate information
  • Focus on a single issue, keep output concise, and solving the problem is sufficient
  • General-purpose design, applicable to fields such as law, business, technology, and academia

Trigger Conditions

Use /multi-search command trigger, or when the user requests:

  • Deeply research multiple related topics
  • Conduct systematic information retrieval
  • Integrate multi-perspective analysis
  • Need to generate a structured research report

Input Format

Method 1: Based on a file

/multi-search @文档路径.md

Method 2: Direct paste

/multi-search
[粘贴材料内容]

Method 3: Specify topics

/multi-search
项目：[项目名称]
研究课题：
1. [课题一]
2. [课题二]
3. [课题三]

Processing Workflow

Stage 1: Analysis Preparation

  1. Read input materials
  2. Extract the list of research topics
    • Has clear topics: use directly
    • No clear topics: automatically extract from materials
  3. Topic splitting principles:
    • Clear direction: Each topic corresponds to a unique retrieval direction
    • Avoid overlap: Ensure retrieval keywords do not repeat across topics
    • Focus on the issue: Each topic solves one specific problem
  4. Determine the project name and output location

Stage 2: Intelligent Detection of Output Directory

Detect the project structure by priority:

  1. Priority detection：output/ directory → use output/[项目名]/
  2. Next detect: current working directory → use ./[项目名]/
  3. Fallback option: user's current directory → use ./research/

Create directory:[输出目录]/03 - 🔍 深度研究/

Stage 3: Parallel Deep Research

Start an independent general-purpose independent research Agent for each research topic.

Context transfer(main Agent → independent research Agent):

  • Key project information (background, goals, core issues)
  • Complete list of topics and the retrieval scope of each topic
  • Assigned keyword directions (the basis for avoiding duplication)
  • Specific requirement background

Deduplication check mechanism：

Before starting retrieval, each independent research Agent must follow this process:

  1. Pre-retrieval declaration：

    • Declare in the current context: "I will retrieve [Keyword A, Keyword B] for researching [Topic Name]"
    • Wait for the main Agent to confirm there is no duplication before starting

  2. Main Agent review：

    • Check whether the keywords declared by that Agent duplicate already assigned directions
    • If duplication is found, promptly notify that Agent to switch to other directions

  3. Dynamic adjustment：

    • If a direction is already covered by another Agent, that Agent should turn to a related but different angle
    • Record the adjusted search direction

In-depth search requirements：

  • 4-6 rounds of in-depth search
  • Automatically select WebSearch (search discovery) or WebFetch (retrieve full content)
  • Differentiate keywords to ensure each Agent covers a unique angle

Document generation：

  • Focus on solving a single core problem
  • Concise and clear, sufficient if it can solve the problem
  • Include key source links
  • Directly usable conclusions and recommendations

Phase 4: Integration and output

  1. Generate research overview document (000.Research Overview.md)
  2. Integrate the core findings of all independent research Agents
  3. Create navigation links between documents
  4. Add comprehensive recommendations and an immediate action checklist

Output format

Directory structure

[输出目录]/
└── [项目名]/
    └── 03 - 🔍 深度研究/
        ├── 000.研究总览.md
        ├── YYMMDD [研究课题一].md
        ├── YYMMDD [研究课题二].md
        └── ...

Overview document format

# [项目名称] 深度研究总览

**生成时间**: YYYY-MM-DD
**研究方式**: N个独立研究 Agent，各进行4-6轮深度检索
**总检索轮次**: XX+轮
**总文档量**: XX KB

---

## 研究成果清单

### 已完成的N份精简研究报告

| 序号 | 研究课题 | 文件大小 | 核心价值 |
|------|---------|---------|---------|
| 01 | [课题一](./YYMMDD%20课题一.md) | XX KB | 简要描述 |

---

## 核心发现

### 发现1：[最重要发现]

**依据**：[简要说明]

**结论**：[具体结论]

---

## 综合建议

### 一、策略建议

**推荐方案**：[具体方案]

### 二、立即行动清单

- [ ] 行动项1
- [ ] 行动项2

Detailed research document format

# [研究课题标题]

**生成时间**: YYYY-MM-DD
**研究深度**: XX+轮深度检索，覆盖XXXX、XXXX、XXXX

---

## 核心结论

[最重要的发现和结论，2-3段，充分详实]

---

## 一、[主要内容一]

### （一）子标题

正文段落。引用来源使用内嵌链接格式：
- 根据[来源名称](https://链接)...
- 依据[资料](https://链接)...

---

## 二、[主要内容二]

[继续结构化内容]

---

## 三、应用建议

### （一）建议要点

**内容**：[具体内容]

### （二）注意事项

⚠️ [注意点]

Link specifications

Core principles

All source links must be embedded in the corresponding position in the body text

✅ 正确：
根据[研究报告](https://链接)显示...

❌ 错误：
根据某报告...
（文末单独列出引用来源）

Link annotation conventions

  • 🔗 → General web resources
  • 📚 → Academic literature
  • 🏛️ → Institutional official websites
  • 📄 → Data sources

Document naming specifications

Numbering system

  • 00. - Research overview
  • 01-09. - Core research
  • 10-19. - Important research
  • 20+. - Extended research

Topic specifications

  • ✅ Use concise Chinese titles
  • ✅ Avoid special characters
  • ✅ Within 15 characters
  • ✅ Clearly reflect the research subject

Quality standards

Independent research Agent research quality

  • Focus on a single problem: Each independent research Agent solves only one core problem
  • Search depth: 4-6 rounds of search (sufficient is enough)
  • Concise output: Concise and clear, sufficient if it can solve the problem
  • Key citations: Cite key sources (sufficient is enough)
  • Directly usable: Provide directly usable conclusions and recommendations

Document quality standards

  • Clear structure: Distinct hierarchy of section headings
  • Coherent narrative: Paragraph-style narrative, avoid excessive listing
  • Accurate links: All links embedded in the corresponding position in the body text
  • Unified format: Follow unified format specifications
  • Strong operability: Provide specific steps, tools, and commands

Notes

Strictly prohibited actions

  • ❌ Do not create second-level subdirectories (such as "reference materials/")
  • ❌ Do not generate a separate execution summary file
  • ❌ Do not use an excessively listed bullet-point format
  • ❌ Do not list citation sources separately at the end of the document
  • ❌ Do not add redundant progress tracking sections

Recommended practices

  • ✅ Use narrative paragraph-style expression
  • ✅ Embed links in the corresponding position in the body text
  • ✅ Keep the research overview concise
  • ✅ Provide specific action recommendations
  • ✅ Mark clear document numbers

Dependencies

This skill relies on Claude Code built-in tools and requires no additional configuration:

  • WebSearch：Search and discover
  • WebFetch：Retrieve full content
  • Task：Launch an independent independent research Agent

Change history

Version Date Update content
v1.0.0 2026-02-15 Migrated from Command to Skill, renamed to multi-topic deep research (multi-search)