Where to Find Tutorials and Guides for AI Berkshire: Complete Documentation Guide

All tutorials and guides for AI Berkshire are built directly into the repository, with the primary documentation located in README.md, the skills/ directory containing 19 skill-specific guides, and the tools/ folder providing utility documentation.

AI Berkshire is a fully-featured, AI-driven investment-research framework designed for Claude Code and Codex. Unlike external documentation sites, every learning resource—from installation steps to advanced research workflows—lives inside the repository itself, making it immediately accessible after cloning.

Installation and Quick-Start Tutorial

The central tutorial entry point is the "快速开始" (Quick Start) section in [README.md](https://github.com/xbtlin/ai-berkshire/blob/main/README.md). This document provides a five-step installation chain:

1. Install the AI client – Choose between Claude Code (npm i -g @anthropic-ai/claude-code) or Codex (npm i -g @openai/codex)
2. Clone the repository – git clone https://github.com/xbtlin/ai-berkshire.git
3. Install the skills – Run the platform-specific script located at scripts/install-claude-commands.sh or scripts/install-codex-skills.sh
4. Verify installation – Confirm with claude --version or codex --version
5. Execute a skill – Test with /investment-research 腾讯 (Claude Code) or investment-research 腾讯 (Codex)

Each step includes concrete command examples and expected output, eliminating the need for external setup guides.

Skill Documentation in the skills/ Directory

The repository ships 19 ready-to-use skills under the skills/ folder. Each Markdown file follows a standardized template containing the research purpose, exact usage syntax, and parameter definitions.

Deep Company Research

The investment-research skill, defined in [skills/investment-research.md](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-research.md), executes a seven-module analysis covering business model, moat evaluation, reverse-thinking, management assessment, civilization trends, valuation, and safety margin calculations.

/investment-research 腾讯

Quality Screening

The quality-screen skill in [skills/quality-screen.md](https://github.com/xbtlin/ai-berkshire/blob/main/skills/quality-screen.md) applies seven hard-filter criteria to ticker lists, returning a clean shortlist for further analysis.

/quality-screen 恒生指数成分股

News Analysis

The news-pulse skill documented in [skills/news-pulse.md](https://github.com/xbtlin/ai-berkshire/blob/main/skills/news-pulse.md) aggregates latest headlines, sentiment scores, and price-impact hypotheses.

/news-pulse 美团

Additional skills such as industry-research, earnings-review, portfolio-review, and thesis-tracker follow identical documentation patterns in the same directory.

Supporting Tools and Utilities

The tools/ directory contains Python utilities referenced by the skills for precise financial operations:

• financial_rigor.py – Handles decimal-based financial calculations and market-cap verification
• stock_screener.py – Batch filters securities against the seven hard criteria
• report_audit.py – Automated sanity-checking of generated Markdown reports
• momentum_backtest.py – Simple back-testing for momentum strategies

These tools include command-line examples in their docstrings and README references, allowing you to invoke them directly when the AI framework requires specific data validation.

Step-by-Step Workflow Example

Below is a complete, reproducible workflow that demonstrates the integrated tutorial experience:

# Clone and install

git clone https://github.com/xbtlin/ai-berkshire.git
cd ai-berkshire
./scripts/install-codex-skills.sh   # or install-claude-commands.sh

# Screen a watchlist

codex /quality-screen "$(cat data/watchlist.json | jq -r '.symbols | join(\",\")')"

# Deep dive on candidate

codex /investment-research 腾讯

# Verify generated report

python3 tools/report_audit.py reports/腾讯/最终报告.md

For Claude Code users, replace codex with claude and remove the leading slash from skill names.

Summary

• Primary tutorials reside in README.md under the "快速开始" section, covering installation through first execution
• Skill-specific guides are the 19 Markdown files in skills/, each providing command syntax and parameter details for research workflows
• Installation automation is handled by scripts/install-claude-commands.sh and scripts/install-codex-skills.sh
• Utility documentation in tools/ explains financial calculations, screening logic, and report validation
• Visual reference is available in assets/architecture.png, showing the three-layer Skill → Agent → Tool design

Frequently Asked Questions

Do I need to visit an external website to learn AI Berkshire?

No. According to the xbtlin/ai-berkshire repository structure, all documentation is self-contained within the repository. The README.md provides the quick-start tutorial, while the skills/ directory contains 19 individual guide files for specific research scenarios.

What is the difference between the Claude Code and Codex installation scripts?

The scripts/install-claude-commands.sh configures commands for Claude Code (using slash-prefixed invocations like /investment-research), while scripts/install-codex-skills.sh sets up the same functionality for Codex (using direct skill names without slashes). Both scripts perform the same underlying skill registration but adapt the syntax to each platform's conventions.

How do I verify that a generated investment report is accurate?

Use the tools/report_audit.py utility to perform automated sanity checks. Run python3 tools/report_audit.py followed by the path to your generated Markdown report (e.g., reports/阿里巴巴/最终报告-投研团队-20260513.md) to validate data consistency and calculation integrity.

Can I extend the framework with my own tutorials?

Yes. The repository includes a CONTRIBUTING.md file that provides guidelines for extending the tutorial set or adding new skills. You can create new Markdown files in the skills/ directory following the established pattern of purpose, usage, and parameters documentation.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}