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

deep-research

Writing & research

View repository

Updated 10-16 01:48

About this skill

Deep Research Skill

Skill Overview

This skill is used to conduct deep research on technical topics and output professional research report documents.

Capability Description
Content Extraction Extract core information from URLs and documents
Deep Research Search online to supplement background, comparisons, and the latest developments
Report Generation By default, generate both Markdown and Word versions
Diagram Generation Generate technical infographics for core concepts
Word Formatting Automatically handle styles such as table of contents, bold headings, and solid table lines

Trigger Rules

Use this skill when the user message contains the following keywords:

  • research, deep research, research report
  • help me research, help me analyze
  • research this, take a look at this

Output Specifications

Each research task must provide both:

  1. Markdown version: used for Obsidian knowledge base accumulation and bidirectional linking
  2. Word version: used for formal reporting and external sharing, and must be processed through script formatting

Directory Structure

Create an independent folder for each research topic to keep things tidy:

{output_dir}/
├── Ralph-Loop/                    # 主题文件夹（英文短横线命名）
│   ├── images/                    # 该主题的信息图
│   │   ├── architecture.png
│   │   └── comparison.png
│   ├── Ralph-Loop调研报告.md      # Markdown 报告
│   └── Ralph-Loop调研报告.docx    # Word 报告
├── MCP-Protocol/
│   ├── images/
│   ├── MCP-Protocol调研报告.md
│   └── MCP-Protocol调研报告.docx
└── ...

Naming conventions:

  • Folder name: English, words connected with hyphens, e.g. Ralph-Loop、MCP-Protocol
  • Report files:{主题名}调研报告.md and {主题名}调研报告.docx
  • Image directory: a separate images/ directory under each topic folder

Research Process

Step 1: Create the topic directory

Create an independent folder based on the research topic:

mkdir -p "{output_dir}/{主题名}/images"

Step 2: Content Acquisition

  1. If the user provides a URL, use webfetch to retrieve the content
  2. Extract core concepts, technical principles, and key information
  3. Identify points that require deeper research

Step 3: Deep Research

Use the Task tool to search online and supplement:

  • Technical background and development history
  • Competitor comparisons and differentiation
  • Community discussions and real-world cases
  • GitHub repositories and open-source implementations
  • Latest developments and trends

Step 4: Diagram Generation

Use preset style scripts to generate unified hand-drawn style infographics.

Image Generation Trigger Rules

Content Type Whether to Generate Image Diagram Type Description
Core architecture/principle Required arch System structure, tech stack, module composition
Process/steps Required flow Workflow, execution order, operation steps
A vs B comparison Required compare Comparison of two solutions/technologies
3 or more elements Recommended concept Core concepts, composed of multiple aspects
Plain text tables Not required - Just use a Markdown table
Code example Not needed - Just use a code block

Preset style templates

All illustrations use a unified hand-drawn visualization style to maintain series consistency:

Type Command parameters Color scheme Layout
Architecture diagram -t arch Tech blue #4A90D9 Layered/modular
Flowchart -t flow Blue + green + orange Top to bottom
Comparison diagram -t compare Blue vs orange Left-right split columns
Concept diagram -t concept Blue-purple gradient Center-radiating

Generation command

Use the research_image.py script to generate:

# 架构图
python .opencode/skills/image-service/scripts/research_image.py \
  -t arch \
  -n "Ralph Loop 核心架构" \
  -c "展示 Prompt、Agent、Stop Hook、Files 四个模块的循环关系" \
  -o "{output_dir}/{主题名}/images/architecture.png"

# 流程图
python .opencode/skills/image-service/scripts/research_image.py \
  -t flow \
  -n "Stop Hook 工作流程" \
  -c "Agent尝试退出、Hook触发、检查条件、允许或阻止退出的完整流程" \
  -o "{output_dir}/{主题名}/images/flow.png"

# 对比图
python .opencode/skills/image-service/scripts/research_image.py \
  -t compare \
  -n "ReAct vs Ralph Loop" \
  -c "左侧ReAct依赖自我评估停止，右侧Ralph使用外部Hook控制" \
  -o "{output_dir}/{主题名}/images/comparison.png"

# 概念图
python .opencode/skills/image-service/scripts/research_image.py \
  -t concept \
  -n "状态持久化要素" \
  -c "中心是Agent，周围是progress.txt、prd.json、Git历史、代码文件" \
  -o "{output_dir}/{主题名}/images/concept.png"

Image naming conventions

Diagram type File name
Architecture diagram architecture.png or {具体名称}_arch.png
Flowchart flow.png or {具体名称}_flow.png
Comparison diagram comparison.png or {A}_vs_{B}.png
Concept diagram concept.png or {具体名称}_concept.png

Step 5: Report writing

Write a Markdown report according to the standard template and save it in the topic folder:

{output_dir}/{主题名}/{主题名}调研报告.md

Reference images in the report using relative paths:

![架构图](images/architecture.png)

Step 6: Word export

# 进入主题目录
cd "{output_dir}/{主题名}"

# 生成 Word（--resource-path=. 确保图片正确引用）
# 注意：不要使用 --toc 参数，因为 Markdown 中已有手写目录
pandoc "{主题名}调研报告.md" -o "{主题名}调研报告.docx" --resource-path=.

# 格式化 Word
python ../../../.opencode/skills/deep-research/scripts/format_docx.py "{主题名}调研报告.docx"

Writing principles

The core value of a research report: in-depth research, reducing the team's cost of absorbing information, and providing expert-level recommendations.

  1. Thorough understanding: no superficial understanding or large-scale copying; must digest and absorb it, then express it in your own words
  2. Show thinking: have judgment and recommendations, rather than merely stating the current state
  3. Support with details: conclusions are supported by process and details, not empty talk
  4. Clear logic: paragraphs, structure, and numbering
  5. Illustration notes: core concepts must be accompanied by infographics
  6. Remove AI flavor:
    • Do not use special symbols such as 「」 or " "
    • Do not use excessive emphasis symbols and emoji
    • Write naturally and fluently, like a professional document written by a human
    • Avoid clichés such as "first, second, in conclusion"

Report template

---
date: YYYY-MM-DD
type: 调研报告
领域: {技术领域}
tags: [调研, {主题关键词}]
---

# XX调研报告

> 调研日期：YYYY年M月D日

---

## 目录

- 一、简介
- 二、启示
- 三、核心介绍
  - 3.1 XXX
  - 3.2 XXX
- 四、附录
  - 4.1 详细文档
  - 4.2 参考资料

---

## 一、简介

（快速说明调研内容，简短重点）

是什么，主要用来做什么，属于什么类别。有哪些能力，有什么特点。和竞品相比，有哪些区别，主打什么。

1. 要点一
2. 要点二
3. 要点三

---

## 二、启示

（调研内容带来的启示、值得学习借鉴之处、与现有产品如何结合、是否值得推荐）

1. 启示一
2. 启示二
3. 启示三

---

## 三、核心介绍

（正文部分，详细说明调研内容的原理/搭建/操作/使用过程，含信息图及流程说明）

### 3.1 XXX

![图解说明](images/xxx.png)

上图展示了...（图解说明，让读者看图就能理解）

详细内容...

### 3.2 XXX

详细内容...

---

## 四、附录

### 4.1 详细文档

（更详细的配置/操作过程）

### 4.2 参考资料

**官方文档**

- 文档名称: https://xxx

**开源实现**

- 项目名称: https://github.com/xxx

**社区讨论**

- 讨论来源: https://xxx

Script description

format_docx.py

Word document formatting script, with functions including:

  1. Centered title, black font (remove pandoc's default blue)
  2. Replace "Table of Contents" with the Chinese "目录"
  3. Separate page for the table of contents
  4. Automatically insert a page break before level-1 headings (Introduction, Insights, etc.)
  5. Keep tables intact without breaking across pages
  6. Keep code blocks intact without breaking
  7. Center the date line

Usage:

python .opencode/skills/deep-research/scripts/format_docx.py "输入.docx" ["输出.docx"]

Complete research example

User input:

Research Ralph Loop

Execution flow:

# 1. 创建主题目录
mkdir -p "{output_dir}/Ralph-Loop/images"

# 2. 获取内容（如有 URL）
webfetch https://example.com/article

# 3. 深度调研（使用 Task 工具联网搜索）

# 4. 生成信息图
python .opencode/skills/image-service/scripts/text_to_image.py "技术架构图..." --output "{output_dir}/Ralph-Loop/images/architecture.png"

# 5. 撰写报告
# 写入 {output_dir}/Ralph-Loop/Ralph-Loop调研报告.md

# 6. 导出 Word（不使用 --toc，Markdown 已有手写目录）
cd "{output_dir}/Ralph-Loop"
pandoc "Ralph-Loop调研报告.md" -o "Ralph-Loop调研报告.docx" --resource-path=.
python ../../../.opencode/skills/deep-research/scripts/format_docx.py "Ralph-Loop调研报告.docx"

Output files:

{output_dir}/Ralph-Loop/
├── images/
│   ├── architecture.png
│   └── comparison.png
├── Ralph-Loop调研报告.md
└── Ralph-Loop调研报告.docx

Dependencies

  • pandoc: Markdown to Word
  • python-docx: Word formatting
  • image-service skill: generate infographics