Private Equity Investment Due Diligence Skill

中文说明在前，English documentation follows.

一个用于私募股权、VC/PE、并购、战略投资、Pre-IPO、成长股权投资等非上市公司投资场景的专业尽职调查 Skill。它把杂乱的 BP、访谈记录、会议纪要、财务表、合同、数据室资料、飞书/云文档和本地底稿，转化为可用于内部投委会讨论的投资尽调报告、IC Memo、红旗扫描、资料清单和管理层访谈提纲。

本 Skill 的默认完整交付物是 standalone HTML 尽调报告，并强制保留基础信息整理、证据边界、资料缺口、红旗登记表和交易影响判断。

目录

• 核心定位
• 适用场景
• 不适用场景
• 能力亮点
• 方法来源
• 工作流
• 默认输出
• HTML 报告要求
• 证据与红旗规则
• 安装方式
• 使用示例
• 仓库结构
• 安全与合规边界
• 维护建议
• English Documentation

核心定位

这个 Skill 不是普通的“报告模板”，而是一个投资尽调执行框架。它要求 Agent 在输出投资观点之前，先完成四件事：

1. 盘点资料边界：区分原始文件、签署版、草稿、会议纪要、AI 转写、管理层口径、第三方报告和未打开附件。
2. 抽取基础信息：主体信息、历史沿革、股权结构、治理、激励、交易条款、产品线、团队、财务基础、证据冲突和资料缺口。
3. 建立证据矩阵：把核心判断拆成 claim -> evidence -> source -> status -> gap -> deal implication。
4. 转化为交易动作：所有重大问题必须落到继续、暂停、重定价、交割条件、陈述保证、赔偿、托管、治理权、投后动作或放弃交易。

适用场景

适合以下任务：

• 私募股权投资尽职调查报告
• VC/成长股权项目初步尽调
• Pre-IPO 项目投资建议书
• 控股收购、并购、少数股权投资、战略投资
• IC Memo / 投委会材料
• 红旗扫描与交易风险判断
• 数据室资料清单和缺口清单
• 管理层、客户、供应商、员工、专家访谈提纲
• 财务尽调 / QoE 初稿
• 估值、回报、退出、交易条款分析
• 投后 100 天计划和价值创造建议

不适用场景

以下场景不建议作为主 Skill 使用：

• A 股、港股、美股等上市公司年报财务排雷
• 二级市场股票买卖建议
• 法律意见书、审计报告、税务意见书的替代
• 需要最终专业签字责任的法律、税务、审计、专利、环保、安全生产判断

对于法律、税务、知识产权、劳动、环保、安全、诉讼和监管问题，本 Skill 只做风险识别和资料请求，不给最终专业意见。

能力亮点

1. 基础信息不会被省略

完整报告必须包含：

• 材料边界和证据状态
• 公司名称、别名、法定代表人、成立时间、经营状态、注册资本、实缴资本
• 历史沿革和重大里程碑
• 投前/投后股权结构
• 控制权、治理、董事/监事/观察员、重大事项保护
• 股权激励、员工持股平台、出资安排
• 交易金额、估值、持股、工具、交割条件、资金用途
• 产品线、产品阶段、商业化进展
• 团队结构、关键人员、能力缺口
• 财务基础、现金、burn rate、收入状态、融资历史
• 源文件冲突和待补资料

2. 五本尽调书逐模块融合

Skill 不是只套用一个 checklist，而是把五类方法合在一起：

3. 默认输出 HTML 报告

完整报告默认生成 standalone HTML，便于本地打开、投委会阅读、后续转 PDF 或归档。

HTML 报告会强调：

• 结论先行
• 表格化事实整理
• 证据状态标记
• 红旗和交易动作
• 移动端/桌面端可读
• 敏感信息不外露

4. 以交易影响为终点

每个红旗必须回答：

• 是否停止交易？
• 是否暂停等待证据？
• 是否需要重定价？
• 是否需要交割条件？
• 是否需要陈述保证、赔偿、托管、回购、反稀释、重大事项保护？
• 是否需要律师、会计师、税务师、专利代理人或行业专家复核？
• 是否转成投后 100 天计划？

方法来源

本 Skill 综合了五本本地尽调资料的方法论：

• 《财务、法律、业务尽职调查全流程实务操作指南》：完整 DD 分类、业务/财务/法律流程、资料清单和报告模板。
• 《财务尽职调查：全流程方法与实务案例》：财务尽调、现场工作、访谈、第三方验证、订单-收入-回款交叉验证、QoE 报告逻辑。
• 《私募股权投资和尽职调查手册》：PE 投资决策、估值、谈判、交易条款和投资报告语言。
• 《投资尽职调查：投资逻辑、尽调方法、实战案例》：价值/估值/退出/风险框架，业务-财务-行业三角验证。
• 《尽职调查白皮书》：经典尽调流程、方法、公司历史、产品、资产、人力、财务、业务、法律、合同、诉讼、环保、IP 清单。

本仓库只包含对方法论的综合抽象和可执行工作流，不包含原书正文。

工作流

完整 DD 报告建议按以下顺序执行：

1. Intake

   • 确认项目名称、交易类型、决策阶段、投资角色、输出格式、资料来源。

2. Material Inventory

   • 盘点 BP/CIM、财务表、合同、数据室、会议纪要、访谈、公开信息、第三方报告。

3. Basic Information Pack

   • 抽取主体基础信息、历史沿革、股权治理、交易事实、产品、团队、财务基础和资料冲突。

4. Investment Thesis

   • 先写清楚“这个交易为什么可能成立”和“必须为真的关键假设”。

5. Adapter Selection

   • 根据交易类型选择 VC、成长股权、Pre-IPO、并购、战略投资等适配器。
   • 根据行业选择硬科技、半导体、SaaS、制造、消费、医药等适配器。

6. Book Fusion Coverage

   • 检查每个模块是否融合了五本书的必要视角。

7. Hypothesis Tree

   • 把投资逻辑拆成业务、行业、财务、法律、团队、估值、退出、投后等可验证问题。

8. Evidence Matrix

   • 把每个重要判断绑定到证据、来源、状态、缺口和交易影响。

9. Triangulation

   • 业务、行业、财务、客户、订单、回款、法律、团队相互校验。

10. Valuation / Return / Exit / Terms

    • 评估估值方法、回报情景、退出路径和条款保护。

11. Red-Flag Gate

    • 按 P0/P1/P2 标记风险，并写明交易动作。

12. HTML Report Draft

    • 生成结论先行、证据可追踪、表格充分的 standalone HTML。

13. Final QA

    • 解析 HTML、扫描敏感内容、检查桌面和移动端可读性。

默认输出

除非用户另有要求，完整报告应包括：

• 执行摘要和投资建议
• 交易背景和资料边界
• 基础信息与历史沿革
• 股权结构、治理、激励和交易条款
• 投资逻辑和关键假设
• 行业与市场尽调
• 产品、技术和商业模式尽调
• 客户、收入、订单、渠道和供应链尽调
• 财务尽调 / QoE
• 法律、税务、合规和交易风险
• 管理层和组织能力
• 估值、回报和退出
• 红旗登记表
• 资料缺口清单
• 交割条件和投后 100 天计划
• 最终建议和决策情景

HTML 报告要求

完整 HTML 报告应满足：

• 单文件 standalone HTML
• inline CSS
• 有封面/hero、目录、章节 ID
• 表格充分，尤其是基础信息、股权、条款、红旗、资料缺口
• 桌面端无页面级横向溢出
• 移动端表格可在 section 内横向滚动
• 无明显文字重叠
• 不暴露原始敏感信息

推荐生成后运行：

python3 - <<'PY'
from html.parser import HTMLParser
from pathlib import Path
p = Path("REPORT.html")
class P(HTMLParser):
    pass
parser = P()
parser.feed(p.read_text(encoding="utf-8"))
print("HTML parser OK", p.stat().st_size)
PY

敏感内容扫描示例：

rg -n '身份证|tenant_access_token|app_ticket|token=|internal-api|user-id|银行账号' REPORT.html || true

证据与红旗规则

证据等级

优先级从高到低：

1. 官方原始文件：审计报告、银行流水、税务申报、合同、发票、验收单、许可证、法院/监管记录。
2. 第三方确认：客户/供应商访谈、银行确认、专家访谈、政府/工商查询、外部律师/会计师报告。
3. 内部系统导出：ERP、CRM、订单台账、仓库、工资、项目系统、生产记录。
4. 管理报表和明细表。
5. 管理层访谈、BP/CIM 口径。
6. 媒体、市场材料、传闻。

管理层口径只能形成假设，不能直接成为结论。

红旗等级

• P0：潜在 deal-stopper 或必须立即升级。
• P1：影响估值、条款、交割条件或专业复核。
• P2：投后监控或整改事项。

每个 P0/P1 必须有交易动作。

安装方式

安装到 Codex Skills

mkdir -p ~/.codex/skills
git clone https://github.com/noahnan-max/private-equity-investment-dd-skill.git ~/.codex/skills/private-equity-investment-dd

如果已经 clone 到其他路径，也可以复制：

rsync -a private-equity-investment-dd-skill/ ~/.codex/skills/private-equity-investment-dd/

验证 Skill 文件

如果本地有 Codex skill-creator 校验脚本：

python3 ~/.codex/skills/.system/skill-creator/scripts/quick_validate.py ~/.codex/skills/private-equity-investment-dd

使用示例

生成完整 HTML 尽调报告

请使用 private-equity-investment-dd skill，基于我提供的 BP、访谈记录、会议纪要、财务表和合同，生成一份完整的 HTML 投资尽调报告。

做红旗扫描

这家公司收入增长很快，但应收账款也快速增长。请基于资料做 PE 项目红旗扫描，输出风险等级、证据、交易影响和下一步查验清单。

做 Pre-IPO 项目 IC Memo

这是一个半导体 Pre-IPO 项目，请基于 BP、审计报告、客户订单和管理层访谈生成 IC Memo，并重点检查收入质量、客户集中度、产能、IP 和上市退出可行性。

补充基础信息

报告里的基础信息太薄了，请重新抽取主体信息、历史沿革、股权结构、治理、交易条款、产品、团队、财务基础和资料缺口。

仓库结构

private-equity-investment-dd-skill/
├── README.md
├── SKILL.md
├── agents/
│   └── openai.yaml
└── references/
    ├── basic-information-extraction.md
    ├── book-fusion-coverage.md
    ├── evidence-and-red-flags.md
    ├── html-report-output.md
    ├── industry-dd-adapters.md
    ├── interview-and-data-room.md
    ├── module-checklists.md
    ├── red-flag-playbook.md
    ├── report-structure.md
    ├── report-writing-examples.md
    ├── self-test.md
    ├── source-books.md
    ├── special-dd-adapters.md
    ├── valuation-and-terms.md
    └── workflow.md

安全与合规边界

不要在用户可见报告中暴露：

• 身份证号、护照号等个人敏感 ID
• 银行账号、签名、印章扫描件
• 私人地址、手机号、邮箱
• Feishu/Lark 内部文档 ID、用户 ID、内部 API URL
• token、app ticket、tenant access token、密钥
• 未脱敏合同原文和保密交易条款

建议写法：

• “源文件包含该编号，报告正文已脱敏。”
• “完整证据见底稿，本文仅保留摘要。”
• “该事项需律师/会计师/专利代理人复核。”

维护建议

后续可以继续扩展：

• 增加更多行业 adapter：新能源、跨境电商、AI 应用、机器人、医疗服务、消费品牌。
• 增加脚本：HTML 报告生成模板、敏感信息扫描器、证据矩阵 CSV 转 HTML。
• 增加案例测试集：早期硬件、SaaS、制造业并购、Pre-IPO、现金流异常项目。
• 增加 Word/PDF 导出流程。

---

English Documentation

Overview

Private Equity Investment Due Diligence Skill is a Codex skill for PE, VC/growth equity, M&A, strategic investment, Pre-IPO, and other non-public-company investment diligence workflows.

It turns messy project materials such as pitch decks, CIMs, management interviews, meeting notes, financial statements, contracts, data-room files, cloud documents, and local working papers into professional investment due diligence reports, IC memos, red-flag scans, data request lists, and interview guides.

For full reports, the default deliverable is a standalone HTML due diligence report with a clear evidence boundary, basic company information, investment judgment, red flags, data gaps, and transaction implications.

When To Use

Use this skill for:

• private equity due diligence reports;
• VC/growth equity preliminary diligence;
• Pre-IPO investment review;
• buyout, M&A, minority investment, and strategic investment;
• IC memos and investment committee materials;
• red-flag scans;
• data-room request lists;
• management/customer/supplier/employee/expert interview guides;
• financial due diligence and QoE-style drafts;
• valuation, return, exit, and term-sheet analysis;
• post-investment 100-day plans.

When Not To Use

This skill is not the primary tool for:

• public-company annual-report red-flag scoring;
• secondary-market stock recommendations;
• final legal opinions, audit reports, tax opinions, or patent opinions;
• professional sign-off on legal, tax, IP, labor, environmental, safety, litigation, or regulatory matters.

For specialist areas, the skill identifies risks and required follow-up. It does not replace lawyers, accountants, tax advisors, patent agents, or other regulated professionals.

What Makes It Different

1. Mandatory Basic Information Extraction

A full report must preserve factual diligence material, including:

• material boundary and evidence status;
• legal entity basics;
• history and milestones;
• pre- and post-transaction cap tables;
• governance, ESOP, shareholder platforms, and investor rights;
• investment amount, valuation, ownership, instrument, closing conditions, and use of proceeds;
• product lines and stage;
• team and organization;
• finance basics such as revenue status, cash, burn rate, debt, and financing runway;
• source conflicts and missing materials.

2. Five-Book Method Fusion

The skill fuses five diligence lenses:

3. HTML-First Full Reports

Full reports default to standalone HTML so they can be opened locally, circulated internally, converted to PDF, or archived.

The HTML report emphasizes:

• conclusion-first writing;
• table-based factual organization;
• evidence status labels;
• red flags and deal actions;
• desktop and mobile readability;
• no exposure of raw sensitive identifiers.

4. Transaction Implication Discipline

Every serious issue must translate into one or more actions:

• stop;
• pause pending evidence;
• reprice;
• condition precedent;
• representation and warranty;
• indemnity;
• escrow or holdback;
• governance right;
• specialist review;
• post-investment action.

Workflow

1. Intake

   • Identify project name, deal type, decision stage, investor role, output format, and material sources.

2. Material Inventory

   • Classify pitch decks, financials, contracts, data-room files, interviews, public information, and third-party reports.

3. Basic Information Pack

   • Extract legal entity facts, milestones, ownership, governance, transaction facts, products, team, finance basics, and source conflicts.

4. Investment Thesis

   • Write what must be true for the deal to work.

5. Adapter Selection

   • Select transaction/stage and industry adapters.

6. Book Fusion Coverage

   • Check whether each module uses the relevant diligence lens.

7. Hypothesis Tree

   • Break the deal thesis into testable business, financial, legal, team, valuation, exit, and post-investment questions.

8. Evidence Matrix

   • Map every important claim to evidence, source, status, gap, and deal implication.

9. Triangulation

   • Cross-check business claims, financial data, industry evidence, customers, orders, collections, legal documents, and team interviews.

10. Valuation / Return / Exit / Terms

    • Review valuation method fit, downside/upside cases, exit feasibility, and terms needed to protect against unresolved risks.

11. Red-Flag Gate

    • Classify P0/P1/P2 issues and define required deal actions.

12. HTML Report Draft

    • Produce a standalone HTML report.

13. Final QA

    • Validate HTML parsing, scan sensitive content, and check desktop/mobile readability.

Default Report Structure

A full report should include:

• executive judgment;
• deal context and material boundary;
• basic company information and milestones;
• equity, governance, incentives, and transaction facts;
• investment thesis and assumptions;
• market and industry diligence;
• product, technology, and business model diligence;
• customer, revenue, order, channel, and supply-chain diligence;
• financial diligence / QoE;
• legal, tax, compliance, and transaction risk;
• management and organization;
• valuation, return, and exit;
• red-flag register;
• data request list;
• closing conditions and post-investment 100-day plan;
• final recommendation and decision scenarios.

HTML Report Requirements

A full HTML report should be:

• a single standalone HTML file;
• styled with inline CSS;
• structured with a cover/hero, table of contents, and section IDs;
• table-heavy for facts, cap tables, terms, red flags, and data gaps;
• readable on desktop and mobile;
• safe from raw sensitive identifiers.

Validation example:

python3 - <<'PY'
from html.parser import HTMLParser
from pathlib import Path
p = Path("REPORT.html")
class P(HTMLParser):
    pass
parser = P()
parser.feed(p.read_text(encoding="utf-8"))
print("HTML parser OK", p.stat().st_size)
PY

Sensitive-content scan example:

rg -n 'ID card|passport|tenant_access_token|app_ticket|token=|internal-api|user-id|bank account' REPORT.html || true

Evidence And Red-Flag Rules

Evidence Hierarchy

Prefer higher-quality evidence:

1. official original documents;
2. third-party confirmations;
3. internal system exports;
4. management accounts and schedules;
5. management interviews or pitch-deck claims;
6. media, marketing materials, or hearsay.

Management claims should create hypotheses, not conclusions.

Red-Flag Severity

• P0: potential deal-stopper or must-escalate issue.
• P1: price, term, condition precedent, or specialist-review issue.
• P2: monitoring or post-investment action.

Every P0/P1 issue must have a deal implication.

Installation

Clone into the Codex skills directory:

mkdir -p ~/.codex/skills
git clone https://github.com/noahnan-max/private-equity-investment-dd-skill.git ~/.codex/skills/private-equity-investment-dd

Or copy the repository:

rsync -a private-equity-investment-dd-skill/ ~/.codex/skills/private-equity-investment-dd/

Validate if the local skill validator is available:

python3 ~/.codex/skills/.system/skill-creator/scripts/quick_validate.py ~/.codex/skills/private-equity-investment-dd

Example Prompts

Full HTML DD Report

Use the private-equity-investment-dd skill to generate a full HTML investment due diligence report based on the pitch deck, interviews, meeting notes, financials, and contracts I provided.

Red-Flag Scan

The company has fast revenue growth, but receivables are also rising quickly. Please run a PE red-flag scan and output severity, evidence, deal implication, and next verification steps.

Pre-IPO IC Memo

This is a semiconductor Pre-IPO deal. Based on the deck, audit report, customer orders, and management interviews, generate an IC memo focused on revenue quality, customer concentration, capacity, IP, and exit feasibility.

Repository Structure

private-equity-investment-dd-skill/
├── README.md
├── SKILL.md
├── agents/
│   └── openai.yaml
└── references/
    ├── basic-information-extraction.md
    ├── book-fusion-coverage.md
    ├── evidence-and-red-flags.md
    ├── html-report-output.md
    ├── industry-dd-adapters.md
    ├── interview-and-data-room.md
    ├── module-checklists.md
    ├── red-flag-playbook.md
    ├── report-structure.md
    ├── report-writing-examples.md
    ├── self-test.md
    ├── source-books.md
    ├── special-dd-adapters.md
    ├── valuation-and-terms.md
    └── workflow.md

Safety Boundary

Do not expose in user-facing reports:

• personal IDs;
• bank accounts, signatures, or stamp scans;
• private addresses, phone numbers, or emails;
• Feishu/Lark internal document IDs or user IDs;
• internal API URLs, tokens, app tickets, tenant access tokens, or secrets;
• unredacted contracts or confidential transaction clauses.

Use summaries or redaction notes instead.

Maintenance Ideas

Potential future extensions:

• more industry adapters such as robotics, AI applications, renewable energy, cross-border ecommerce, and healthcare services;
• scripts for HTML report generation, sensitive-content scanning, and evidence-matrix conversion;
• case test sets for early-stage hardware, SaaS, manufacturing M&A, Pre-IPO, and cash-flow-stress deals;
• Word/PDF export workflows.