Guide · Agent skill
How to use the Buffett-Style Pre-Buy Checklist agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The investment-checklist skill is a research aid that tells an AI agent to screen one or more companies against a six-stage value-investing checklist modelled on Warren Buffett's approach. It is designed for individual investors, students and anyone comparing several listed companies in one table. After this guide you will know what the skill does, how to install it, how to trigger it with a plain request, and what it needs to run.
Buffett-Style Pre-Buy Checklistxbtlin/ai-berkshire · MIT Download the skill Quick start
- 1Download the ZIP. Download the investment-checklist ZIP from this page. It contains SKILL.md.
- 2Install it. Unzip it into ~/.claude/skills/ for all projects or into .claude/skills/ inside one project. Alternatively, upload the ZIP in the Claude apps under Settings, Capabilities, Skills.
- 3Get the repository. Make sure you have a checkout of the ai-berkshire repository with AGENTS.md and tools/financial_rigor.py, since the skill refers to them.
- 4Ask for an analysis. Request a checklist in plain language, naming one or more companies or tickers.
- 5Review the report. Open the Markdown file saved to your home folder and check the stated data cutoff date and the sources behind each figure.
What the skill makes the agent do
Once installed, you give the agent one or more company names or tickers. The agent works through six checks: whether the business is understandable, whether it is a good business (returns, margins, cash flow), how wide its moat is, whether management is trustworthy, whether the price leaves a safety margin, and whether your emotions are under control.
The skill gathers company data from the web in parallel, scores each stage from one to five stars, runs some valuation maths through a bundled Python script, and saves a Markdown report to your home folder. The source describes the output as a filter for weak candidates, not as buy or sell advice.
- Stages: circle of competence, economic characteristics, moat, management, margin of safety, decision discipline
- Output: a per-company chapter and, when you analyse several companies, a summary comparison table
- Report file: written to your home folder under a name based on the company or 'multi-company comparison'
Install the skill from the ZIP or the Claude apps
Option one: download the ZIP from this page and unzip it into your user skills folder at ~/.claude/skills/, or into a project's .claude/skills/ folder if you want the skill available only to that project. Option two: in the Claude apps, open Settings, go to Capabilities, then Skills, and upload the ZIP.
The skill package contains a single file, SKILL.md. The skill itself refers to the ai-berkshire repository, so read the requirements section below before expecting every step to work from the ZIP alone.
- Upload path in the apps: Settings → Capabilities → Skills
- Local folders: ~/.claude/skills/ for all projects, or .claude/skills/ inside one project
Check the requirements before you run it
The skill expects a checkout of the ai-berkshire repository. It refers to AGENTS.md and to a shared tools/ folder, and it calls tools/financial_rigor.py with Python 3. The skill is written to run commands from the repository root, so the script path must resolve from there.
You also need shell access so the agent can run the date and python3 commands, and web search so it can gather company data. The skill's files reference Task, Agent and WebSearch tools; on other platforms the agent is told to use the nearest equivalent available in that session.
- Python 3
- The ai-berkshire repository checkout, including AGENTS.md and tools/financial_rigor.py
- Shell access for date and python3
- Web search capability
Trigger the skill with a natural request
You do not need special syntax. Name the companies and ask for a pre-buy checklist. The skill accepts several companies separated by commas, Chinese enumeration commas, or spaces, and it accepts names or ticker symbols.
The skill begins by running the date command, so it can state the data cutoff date in the report header. It then parses your companies and checks each one's listing status. Unlisted companies are marked as such and skipped for the full checklist.
- Ask: Run the Buffett pre-buy checklist on Nvidia, Apple and Microsoft
- Ask: Use the investment checklist on NVDA AAPL MSFT and give me a comparison table
- Ask: Screen Moutai and Tencent with the value investing checklist and save the report
Read the output and what each score means
Each listed company gets a six-stage table with scores from one to five stars, a core data table, three to five key risks, a mirror test, and a clear result. The result is one of four labels: passed, not passed (with the triggered red line named), grey zone (the agent lists the open questions for you), or not applicable.
The mirror test asks the agent to write five sentences explaining a buy decision. If it cannot complete all five, the skill marks the company as not to buy. Treat the stars as subjective judgements that organise your thinking, not as measurements.
- Passed: can proceed to deep research
- Not passed: a hard veto was triggered
- Grey zone: key disputes need your own judgement
- Not applicable: unlisted or cannot be bought
Safety notes: scripts and network access
This skill runs scripts and uses the network. The valuation steps call tools/financial_rigor.py through python3, with arguments for price, earnings per share, shares, growth and P/E scenarios. The data gathering step uses web search and sub-agents, so company information is fetched from online sources.
The source does not name the sources the agent uses for prices or financials. Figures can therefore be stale or wrong. Cross-check every number against primary filings and a current quote before relying on it, and read the script before you run the skill on a machine that holds sensitive files.
- Runs local Python scripts: yes
- Uses network access: yes
- Data sources are not named in the skill, so verify key figures yourself
python3 tools/financial_rigor.py three-scenario --price {price} --eps {EPS} --shares {shares} --growth {optimistic} {neutral} {pessimistic} --pe {optimistic PE} {neutral PE} {pessimistic PE} --currency {currency}What people use it for
- Filtering a watchlist of listed companies before reading their annual reports
- Comparing three or four companies side by side in one summary table
- Studying how value-investing criteria such as moat, management and margin of safety apply to real businesses
- Documenting a written buy thesis and testing whether it can be stated in five sentences
Frequently asked questions
Is the investment checklist skill free?
The source gives the licence as MIT, which generally permits use, copying and modification with the licence notice retained. The source does not state any price, so check the repository licence file for the exact terms.
What do I need to run the investment-checklist skill?
You need Python 3, a checkout of the ai-berkshire repository including AGENTS.md and tools/financial_rigor.py, shell access for date and python3 commands, and a web search capability. The skill is written for Claude Code and Codex style agents.
Where do I install an agent skill like this?
Unzip the skill into ~/.claude/skills/ for all projects, or into .claude/skills/ inside a single project. In the Claude apps you can upload the ZIP under Settings, Capabilities, Skills.
Does the skill give buy or sell recommendations?
No. The source describes it as a research aid for filtering out weak candidates, not a source of buy or sell advice. Its scores are subjective judgements, so you still need independent research before any trade.
Can it analyse unlisted companies or private investments?
No, not in full. The skill marks unlisted companies as not analysable and skips the complete checklist for them, though it gives a brief note on whether there is any indirect route to invest.
How does it compare with simply asking an AI chatbot about a stock?
The skill imposes a fixed six-stage structure, a mirror test and a list of hard vetoes, and it saves a report file. A general chat answer has none of this fixed framework. Neither replaces reading primary filings, and the skill's sources are not named in its files.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you want a disciplined, repeatable first-pass screen and you are comfortable setting up the ai-berkshire repository and cross-checking figures yourself. Pick something else if you need reliable real-time prices, verified financials, or analysis of private companies, because the skill does not guarantee any of these.
Download the skill