AI Berkshire

Install

Manual install

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

This is third-party code. Review the repository files before installing.

What it does

The set gives the agent about twenty skills for systematic analysis of public and private companies. It builds on the approaches of Warren Buffett, Charlie Munger, Duan Yongping and Li Lu, and for each company it assembles a review of the business model, financials and valuation, industry structure, risks and management quality. Separate skills cover deep research, earnings review, industry screening, thesis tracking and quick attribution of a price move. The skill requires naming data sources, cross-checking figures against two independent sources and flagging uncertainty, and it frames the result as analysis with ranges and caveats rather than a buy or sell recommendation. It runs as slash commands in Claude Code and as SKILL.md packages in Codex from one shared rule set.

Who it is for. For investors, analysts and founders who need a disciplined company review rather than a ready answer.

Example request

Analyze this company across business model, financials, industry and risk and show valuation ranges, without a buy call

Limitations

This is a research tool, not investment advice: decisions and their consequences stay with the user, and past returns do not guarantee future ones. Deep skills run several passes and cross-checks and consume many tokens, and the author suggests a strong model for important decisions. An installed Claude Code or Codex client and access to their models are required. Analysis quality depends on how much public data a company has, and several skills run helper scripts and fetch financial data from the web. The material is mainly in Chinese, with English and Japanese README files.

How to disable. Remove the installed slash commands from the Claude Code commands directory or the skill folders from ~/.codex/skills, depending on how you installed them.

Security check

• Skills run helper scripts and fetch financial data from the web
• The topic involves money: outputs are analysis, not advice, and decisions remain the user's responsibility

README in short

The README describes a set of about twenty investment research skills compatible with Claude Code and Codex. The material systematizes the methodologies of Buffett, Munger, Duan Yongping and Li Lu and requires naming sources, cross-checking financial data and flagging uncertainty. The repository keeps one shared rule set: slash commands for Claude Code in skills/, SKILL.md packages for Codex in codex-skills/ and an optional slash-prompt layer in codex-prompts/. Installation goes through git clone and the install-claude-commands or install-codex-skills scripts. The repository ships example reports and data, under the MIT license.

SKILL.md

---
name: investment-research
description: "AI Berkshire skill: 投资研究：巴菲特-芒格-段永平-李录 四大师综合分析框架. Source: skills/investment-research.md."
---

## Codex adapter note

This skill is generated from `skills/investment-research.md` so Claude Code and Codex users share one canonical workflow.

- Treat `$ARGUMENTS` as the user's request in the current Codex thread.
- When the source mentions Claude-only surfaces such as Task, Agent, WebSearch, Bash, Read, or Write, use the closest Codex capability available in this session: subagents when available, web search when needed, shell commands for local tools, and normal file edits for workspace files.
- Use shared project tools from `tools/` in this repository. Prefer running commands from the repository root with paths like `python3 tools/financial_rigor.py ...`; if the current thread starts outside the repo, locate the actual checkout path first instead of assuming a fixed home-directory path.
- Before starting research, run the `date` command to confirm today's date; treat it as the baseline for "latest" data and state the data cutoff date in the report header. Never assume the current date from training data.
- Preserve the research quality rules from `AGENTS.md`: cross-check financial data, use exact arithmetic tools for valuation/math, and clearly label uncertainty and source gaps.

Related