🔎
AI Engineering

Deep Research

by SeanEllyJames

Deep Research is an AI Engineering skill for Claude Code, published by SeanEllyJames in deep-research-skill.

201 stars0 forkson SeanEllyJames/deep-research-skillAdded 2026/09/05+14% in starsRepository updated 2026/08/18
aiai-agentclaudeclaude-codedeep-researchllmprompt-engineeringresearchskill
Install in seconds
Install Deep Research
Copy Deep Research into your Claude Code skills folder. Run the command in your terminal, or review the source on GitHub before installing.
terminal
npx degit https://github.com/SeanEllyJames/deep-research-skill ~/.claude/skills/deep-research-skill

Requires Node.js. Downloads this skill only — not the rest of the repository — into your Claude Code skills folder.

Without Node.js

git clone https://github.com/SeanEllyJames/deep-research-skill.git

Clones the whole repository, then copy the skill’s own directory into your skills folder yourself.

In this catalog

Source file
SKILL.md in SeanEllyJames/deep-research-skill
Installs to
~/.claude/skills/deep-research-skill
Collection
The only skill cataloged from this repository
Category
AI Engineering3101 skills

What Deep Research does

Deep Research structures a Claude Code research run around primary sources, comparison, validation, and a single final report. Use it when you need a defensible judgment instead of a source roundup.

Deep Research is cataloged under AI Engineering on DirSkills. Deep Research comes from a repository tagged ai, ai-agent, claude, claude-code and deep-research.

Documentation

README

深度调研工作流

🚦 硬上限:任意时刻并行 subagent ≤7 个。 维度多就分批,不要抬高上限。子 agent 禁止再派子 agent——递归 fan-out 是最烧钱的失败模式(见错误 6)。如果你的平台自带一个 fan-out 到几十个 agent 的"deep research"模式,不要用它,用本流程做受控并行。

元数据

  • 类型: Workflow
  • 适用场景: 需要对某个主题进行深度、全面、可验证的第三方调研,并且你要的是一个"判断",不是一份综述
  • 输出位置: 你指定的一个输出目录

核心原则

  1. 一手源优先: 先找原始文献(官方公告、技术论文、创始人博文、财报电话会议),精读后再做广度搜索。二手信息层层转手会蒸发洞见
  2. 论点驱动: 报告的结构是论证链(观察 → 分析 → 判断),不是主题分类(市场 → 竞争 → 趋势)。每一章推进一个论点,前后有逻辑递进
  3. 交叉对比: 选 2-3 个核心案例做深度对比,找收敛点和分歧点。收敛本身就是洞见,分歧背后的原因更是洞见
  4. 标注不确定性: 诚实区分"确定知道的"、"合理推断的"、"不确定的"。过于确定的结论是分析不严谨的信号
  5. 可追溯性: 所有引用必须保留 URL,关键引用保留原文摘录
  6. 单一交付: 最终只交付一个汇总报告,中间结果不保存

关键区分:Wide Research vs Deep Research

Wide Research(信息搬运):搜索多个来源 → 按主题分类摆放 → 输出 checklist。这是 LLM 的默认模式,产出的是"正确的废话"。

This is the opening of the README. Read the full README on GitHub.

Frequently asked about Deep Research

  • How does Deep Research compare to other AI Engineering skills?

    Deep Research ranks #2844 by stars among the 3101 AI Engineering skills in this catalog. The most-starred ones next to it are Architecture Decision Records, AI-First Engineering and Agentic OS. DirSkills ranks by the star count of the repository each skill ships in, so that order reflects how popular those repositories are rather than any review of Deep Research against them. Open each page to compare what they document and how they install.