Skip to content

zhangswings/work-skill

BranchesTags

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
deep-research-product
deep-research-product
 
 
examples
examples
 
 
.gitignore
.gitignore
 
 
README.md
README.md
 
 

Repository files navigation

English | 中文

Deep Research Product · 深度产品研究 Skill

一个 Trae IDE / Claude Code / Cursor Skill，用于对产品或公司进行结构化的多模块深度研究，最终输出为单一 Markdown 文件，包含经过事实核查、以一手信息源为驱动的 8 大研究模块。

适用于竞品分析、投资研究、产品调研、行业深度报告等场景。自适应产品类型（SaaS、AI 原生、开源工具、消费品、连锁品牌等）。

效果

  • 📊 8 大研究模块：创始人与哲学、时间线与商业模式、增长引擎分析、用例图谱、生态扩展性、社区与KOL网络、竞争格局、风险与天花板
  • 🔍 产品类型自适应：Phase 0 自动分类产品类型，后续模块针对性调整研究框架和信息源
  • 📈 增长引擎组合分析：不只看 PLG 或 Sales-led 的标签，而是诊断产品实际的多引擎组合（PLG / 付费获客 / KOL / 内容SEO / 销售驱动 / 平台分发 / 网络效应 / 社区 / 渠道合作）
  • ✅ 一手信息源驱动：创始人博客 > 官网 > 论坛帖子 > 播客访谈 > HN 讨论，严格区分一手/二手来源
  • 📝 Obsidian 兼容 Markdown：支持 wikilink、callout、frontmatter，可直接作为 Obsidian Vault 使用
  • 📄 单一文件交付物：所有模块编译为一个 .md 文件，方便分享和阅读

适合 / 不适合

✅ 适合：竞品深度分析 / 投资标的调研 / 产品战略研究 / 行业深度报告 / 了解一个公司/产品的全貌

❌ 不适合：快速的简单问题 / 单点信息查询 / 实时数据监控 / 需要付费数据库支持的量化研究

安装

方式一：把下面这段话直接发给 AI（推荐）

帮我安装 deep-research-product 这个 Skill。请按下面步骤做：

  1. 确保工作目录下的 skills/ 目录存在（不存在就创建）
  2. 执行 git clone https://github.com/zhangswings/work-skill.git
  3. 将 work-skill/deep-research-product/ 目录复制到你的 Skill 目录
  4. 验证：确认目录下有 SKILL.md、references/ 两项
  5. 告诉我安装好了，之后我说"调研 XXX"之类的话就会触发这个 Skill

方式二：手动命令行

git clone https://github.com/zhangswings/work-skill.git

触发方式

装好后，AI Agent 会在对话里自动发现并调用这个 Skill。触发关键词：

  • "调研 [产品名]"
  • "深度研究 [产品名]"
  • "研究一下 [产品名]"
  • "research [product name]"
  • "deep dive on [product name]"
  • "competitive analysis on [product name]"
  • "I want to understand how [X] grew"

使用流程

Skill 本身是结构化的 4 阶段工作流：

  1. Phase 0：产品分类 — 沿3个维度（资金与出身、增长引擎、扩展性）诊断产品类型，生成路由配置
  2. Phase 1：确认上下文 — 确认产品名称、范围、输出语言、可选的"你的产品"视角
  3. Phase 2-3：规划并执行模块 — 根据 Phase 0 路由选择模块集，分批并行执行，每模块独立保存中间文件
  4. Phase 4：编译交付物 — 事实核查关键指标 → 生成 MOC → 可选的综合分析 → 编译为单一 .md 文件

详细说明见 SKILL.md。

研究模块

模块 描述 运行条件
1 · 创始人与哲学 创始人背景、核心哲学原则、决策驱动力 始终运行
2-3 · 时间线与商业模式 关键里程碑、定价矩阵、收入结构、定价哲学 始终运行
4 · 增长引擎分析 诊断实际增长引擎组合、交叉信号分析 始终运行（最重模块）
4b · 用例图谱 核心用户场景、使用频率、替代品分析 始终运行
5 · 生态扩展性 插件/API/SDK/市场生态 扩展性 ≠ None 时运行
6 · 社区与KOL网络 社区形态、KOL分层、口碑走向 始终运行
7 · 竞争格局 竞品矩阵、护城河分析、威胁评估 始终运行
8 · 风险与天花板 核心风险矩阵、天花板分析、生存概率 始终运行

产品分类维度（Phase 0）

维度 选项 影响
资金与出身 Indie / Bootstrapped / VC-backed / Big-tech-internal / Public 模块1深度、模块8风险类别
增长引擎 组合模型 — 对每个引擎评分0-3 模块4深度分配
扩展性 Plugin / API / SDK / MCP / Marketplace / None 模块5运行/跳过

目录结构

work-skill/
├── README.md                       ← 本文件
├── .gitignore
├── deep-research-product/
│   ├── SKILL.md                    ← Skill 主文件：工作流、原则、分类规则
│   └── references/
│       ├── module-prompts.md       ← 8 个模块的参数化提示词模板
│       └── output-format.md        ← Obsidian 格式规范（frontmatter / wikilink / callout）
└── examples/
    ├── luckin-cotti-research.md    ← 瑞幸咖啡 vs 库迪咖啡竞品研究案例
    └── starbucks-research.md       ← 星巴克深度研究案例

硬规则（核心原则）

  1. 挑战假设 — 用户提供的信息常常部分错误（创始人信息、日期、价格），必须用一手来源验证
  2. 一手来源优于二手 — 创始人博客 > 官网 > 论坛帖子 > 播客访谈 > HN讨论 >> Reddit摘要
  3. 真实引语，标明出处 — 每个哲学/痛点声明必须附原文引语+URL
  4. 诚实面对信息空白 — 来源不可达时明确说明，绝不编造
  5. 不向子代理注入未验证数据 — 编排者的一手信息可能过时，关键指标标注"待验证"

输出规范

  • Obsidian 风格 Markdown：每个文件带 YAML frontmatter
  • Wikilinks：[[01-founders-philosophy]]、[[Person Name]]
  • Callouts：> [!info]、> [!quote]、> [!important]、> [!warning]、> [!danger]、> [!tip]
  • Sources 区块：每个文件末尾附来源列表，区分一手/二手/不可达
  • 字数预算：中文每模块 1500-5500 字，英文约 3 倍

案例

  • 瑞幸咖啡 vs 库迪咖啡 — 中国咖啡连锁竞品深度研究，覆盖两家品牌从创立到 2026 年的完整竞争轨迹
  • 星巴克 — 全球咖啡帝国深度研究，聚焦"第三空间"哲学、博裕资本中国合资转型、"Back to Starbucks"战略复苏

License

MIT

Deep Research Product · Deep Product Research Skill

A Trae IDE / Claude Code / Cursor Skill for structured, multi-module deep research on any product or company. The final deliverable is a single Markdown file containing fact-checked, primary-source-driven research across 8 modules.

Use it for competitive analysis, investment research, product strategy, industry deep dives, or understanding how a company/product grew. Adapts to product type (SaaS, AI-native, OSS dev tools, consumer products, retail chains, etc.).

Features

  • 📊 8 Research Modules: Founders & Philosophy, Timeline & Business Model, Growth Engine, Use Case Map, Extensibility & Ecosystem, Community & KOL Network, Competitive Landscape, Risks & Ceiling
  • 🔍 Product Type Classification: Phase 0 auto-classifies the product, subsequent modules adapt research framing and source mix
  • 📈 Growth Engine Composition Analysis: Diagnoses the actual multi-engine combination (PLG / Paid / KOL / Content-SEO / Sales-led / Platform / Network / Community / Channel), not just the marketing label
  • ✅ Primary-Source Driven: Founder blogs > official site > forum posts > podcast interviews > HN threads; strict primary/secondary distinction
  • 📝 Obsidian-Compatible Markdown: Supports wikilinks, callouts, frontmatter — works as an Obsidian Vault out of the box
  • 📄 Single-File Deliverable: All modules compiled into one .md file for easy sharing

Suitable / Not Suitable

✅ Suitable: Competitive deep dives / Investment research / Product strategy / Industry reports / Understanding a company's full story

❌ Not Suitable: Quick one-off questions / Single data point lookups / Real-time monitoring / Quantitative research requiring paid databases

Installation

Option 1: Paste this to your AI agent (Recommended)

Install the deep-research-product skill. Steps:

  1. Ensure skills/ directory exists in the working directory
  2. Run git clone https://github.com/zhangswings/work-skill.git
  3. Copy work-skill/deep-research-product/ to your skill directory
  4. Verify: confirm SKILL.md and references/ exist in the directory
  5. Tell me it's installed. From now on, "research [X]" will trigger this skill.

Option 2: Command line

git clone https://github.com/zhangswings/work-skill.git

Triggers

Once installed, the AI agent will discover and invoke this skill automatically. Trigger keywords:

  • "调研 [product name]"
  • "深度研究 [product name]"
  • "research [product name]"
  • "deep dive on [product name]"
  • "competitive analysis on [product name]"
  • "I want to understand how [X] grew"

Workflow

The skill follows a 4-phase structured workflow:

  1. Phase 0: Classify Product — Diagnose product along 3 dimensions (funding, growth engine, extensibility), generate routing config
  2. Phase 1: Confirm Context — Confirm product name, scope, output language, optional "your product" lens
  3. Phase 2-3: Plan & Execute Modules — Select module set based on Phase 0 routing, run in parallel batches, save intermediate files
  4. Phase 4: Compile Deliverable — Fact-check key metrics → Generate MOC → Optional synthesis → Compile single .md file

See SKILL.md for details.

Research Modules

Module Description Run When
1 · Founders & Philosophy Founder background, core principles, decision drivers Always
2-3 · Timeline & Business Model Key milestones, pricing matrix, revenue streams Always
4 · Growth Engine Diagnose actual growth engine composition, cross-signal analysis Always (heaviest module)
4b · Use Case Map Core user scenarios, frequency, substitutes Always
5 · Extensibility & Ecosystem Plugins/API/SDK/marketplace ecosystem Extensibility ≠ None
6 · Community & KOL Network Community shape, KOL tiers, sentiment trends Always
7 · Competitive Landscape Competitor matrix, moat analysis, threat assessment Always
8 · Risks & Ceiling Risk matrix, ceiling analysis, survival probability Always

Directory Structure

work-skill/
├── README.md                       ← This file
├── .gitignore
├── deep-research-product/
│   ├── SKILL.md                    ← Skill definition: workflow, principles, classification rules
│   └── references/
│       ├── module-prompts.md       ← 8 parameterized module prompt templates
│       └── output-format.md        ← Obsidian formatting conventions
└── examples/
    ├── luckin-cotti-research.md    ← Luckin vs Cotti competitive research case
    └── starbucks-research.md       ← Starbucks deep research case

Hard Rules (Core Principles)

  1. Challenge assumptions — User-provided info is often partially wrong; verify against primary sources
  2. Primary sources beat secondary — Founder blogs > official site > forum posts > podcasts > HN threads >> Reddit summaries
  3. Real quotes, attributed — Every philosophical claim must have original-language quote + URL
  4. Honest about gaps — Explicitly state when sources are unreachable; never fabricate
  5. Don't pre-fill unverified metrics — Mark key metrics as "unverified" for sub-agents

Output Conventions

  • Obsidian-style Markdown: YAML frontmatter on every file
  • Wikilinks: [[01-founders-philosophy]], [[Person Name]]
  • Callouts: > [!info], > [!quote], > [!important], > [!warning], > [!danger], > [!tip]
  • Sources section: Every file ends with source list, categorized as primary/secondary/unreachable
  • Length budgets: Chinese 1,500-5,500 chars per module; English ~3x

Examples

  • Luckin Coffee vs Cotti Coffee — Chinese coffee chain competitive deep dive, covering the complete rivalry from founding to 2026
  • Starbucks — Global coffee empire deep dive, focusing on "Third Place" philosophy, Boyu Capital China JV transition, and "Back to Starbucks" turnaround

License

MIT

About

Deep research product skill for comprehensive product/company analysis

Resources

Readme
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

Packages

Contributors