Guide · Agent skill
How to use the Investment Research Framework agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The investment-research skill is a set of written instructions that tells an AI agent, such as Claude or Codex, how to research a listed company in a fixed order and produce a checklist-style investment report. This guide explains what the skill makes the agent do, how to install it from the ZIP on our page, how to trigger it, and what it needs to run. By the end you will be able to set it up, ask for a report, and judge how much weight the output deserves.
Investment Research Frameworkxbtlin/ai-berkshire · MIT Download the skill Quick start
- 1Get the repository. Obtain the full ai-berkshire repository so the tools/ folder and supporting files are present, not just SKILL.md.
- 2Install the skill. Unzip the downloaded ZIP into ~/.claude/skills/ for all projects, or into a project's .claude/skills/ folder, or upload it under Settings, Capabilities, Skills in the Claude apps.
- 3Check Python 3 and shell access. Confirm Python 3 runs from the terminal and that the agent can run shell commands, since the skill calls the date command and the Python tools.
- 4Start from the repository root. Open the agent in the repository root so the paths like tools/financial_rigor.py resolve correctly.
- 5Request the research. Name the listed company and its market, and ask for a full research report with the data cutoff date stated.
- 6Review the cross-checks. Look at the verification output and the source differences the agent reports, and confirm any figure that matters before you rely on it.
What the skill makes the agent do
The skill is named investment-research and is based on the AI Berkshire repository. Its description refers to a combined framework drawing on the methods of Buffett, Munger, Duan Yongping and Li Lu. In practice, the agent works through seven modules in a fixed order: data collection, business quality, moat, risk review, management, industry trends, and valuation.
Before collecting data, the agent is told to rate how well an AI can research the company, from A (heavily covered) to C (scarce information), and to watch for bias. It is asked to date-stamp its data, cross-check key figures against at least two sources, and run calculations through bundled Python tools instead of estimating in its head. The result is a written report that labels uncertainty and states how much weight the conclusion deserves.
- Seven research modules run in order
- Figures are cross-checked against at least two sources, with differences above 1% flagged
- Calculations go through bundled Python tools
- The output labels uncertainty and source gaps
Install the skill from the ZIP
Download the ZIP for this skill from our page and unzip it. To make it available to Claude Code for all your projects, place the folder in ~/.claude/skills/. To limit it to one project, place it in that project's .claude/skills/ folder.
In the Claude apps, you can instead upload the skill under Settings, then Capabilities, then Skills.
The ZIP contains only SKILL.md. The skill's instructions refer to a tools/ folder and an AGENTS.md file that are not in that file, so you also need the full ai-berkshire repository (see requirements below).
- Claude Code, all projects: ~/.claude/skills/
- Claude Code, one project: .claude/skills/ inside that project
- Claude apps: Settings, then Capabilities, then Skills, then upload
Requirements before you run it
The skill needs more than the SKILL.md file. It runs local Python scripts from the repository, so you need Python 3 and a checkout of the full ai-berkshire repository. The agent also needs shell access to run the date command and its Bash tools, and web access to collect financial data.
The skill names specific sources to use: macrotrends, stockanalysis, AASTOCKS, Eastmoney and CNINFO. Which sources you can reach depends on your setup and network.
- Python 3, to run tools/financial_rigor.py and tools/terminal_value.py
- The full ai-berkshire repository, not just SKILL.md
- Shell access for the date command and Bash tools
- Web search or web access to collect financial data
dateTrigger the skill with a natural request
The skill takes your request as its input, so you trigger it by naming the company you want researched and asking for a full investment research report. Run the agent from the repository root so the tool paths resolve. The skill tells the agent to run the date command first and to state the data cutoff date in the report header.
A typical request names the company and the market, for example: use the investment-research skill to write a research report on a specific listed company, and say whether it is listed in the US, Hong Kong or mainland China so the agent picks the right data sources. Exact wording is flexible; the skill is designed around a company name plus a request for research.
- Name the company and its listing market (US, Hong Kong or A-shares)
- Ask for a full written report, not a quick opinion
- Ask the agent to state the data cutoff date and label uncertainty
Check the numbers with the bundled tools
The valuation step relies on Python tools so the arithmetic is exact. The agent is instructed to verify market capitalisation by multiplying price by total shares, to cross-validate revenue, net profit and cash reserves across sources, and to verify valuation ratios. You can run these checks yourself to confirm the agent's work.
The skill also includes a long-term discounted valuation step using tools/terminal_value.py. The source says this model is a stress test for position sizing, not a buy, hold or avoid decision on its own.
- Market cap check: financial_rigor.py verify-market-cap
- Cross-validation: financial_rigor.py cross-validate
- Valuation ratios: financial_rigor.py verify-valuation
- Three-year scenarios: financial_rigor.py three-scenario
- Long-term model: terminal_value.py audit, pe, irr and flip
python3 tools/financial_rigor.py verify-market-cap \
--price {股价} --shares {总股本} --reported {报告市值} --currency {币种}
python3 tools/financial_rigor.py cross-validate \
--field {字段名} --values '{"来源1": 数值, "来源2": 数值}' --unit {单位}
python3 tools/financial_rigor.py verify-valuation \
--price {股价} --eps {EPS} --bvps {每股净资产} --fcf-per-share {每股FCF} --dividend {每股股息}Read the report with the right expectations
The skill is built to resist overconfidence. It warns that AI research can echo market consensus, and that it may fill data gaps with plausible guesses. For companies with little public coverage, it tells the agent to reason from the business itself rather than assemble a report that looks complete.
The report separates the AI research confidence from the actual investment certainty. Treat the output as a structured draft to check, not as a final answer.
- Heavily covered stocks risk reflecting consensus pricing
- Thinly covered stocks risk being judged 'unclear = bad' by the agent
- Verify key figures yourself before acting
What people use it for
- Building a seven-module written report on a listed company for a study group or investment club
- Cross-checking market cap, revenue and cash figures before writing an analyst note
- Listing the ways a company could fail and the bull and bear arguments in one structured document
- Stress-testing a position size with the long-term discounted valuation step
Frequently asked questions
Is the investment-research skill free to use?
The source material reviewed for this page lists the MIT licence for the codex-skills/investment-research entry in the ai-berkshire repository. Check the repository's licence file for the full terms before reuse.
What do I need to install the investment-research skill?
You need the skill's ZIP unzipped into ~/.claude/skills/ or a project's .claude/skills/ folder, or uploaded in the Claude apps. You also need Python 3 and the full ai-berkshire repository, because the skill refers to tools/ and AGENTS.md files that are not in SKILL.md alone.
Where do I put the investment-research skill folder in Claude Code?
Place it in ~/.claude/skills/ to make it available across projects, or in .claude/skills/ inside a single project to limit it to that project.
Does the skill give real-time market data?
No. The skill relies on web collection and the agent's own sources, which may be incomplete or wrong. The skill itself tells users to verify the output independently.
Can I use this skill as personalised financial advice?
No. It is described as a research workflow, not a licensed advisory service, and it is not designed for buy or sell decisions without independent verification.
How does this compare with asking an AI for a stock opinion?
A plain request may give a quick opinion. This skill forces a fixed seven-step order, date-stamped data, multi-source cross-checks and exact calculations through Python tools, so the output is a structured report that labels its uncertainty.
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
This skill suits investors, students and analysts who want a disciplined, checklist-style research report and are willing to check the AI's figures critically. It is a poor fit for anyone who needs real-time data, personalised advice, or a ready-made buy or sell decision, and for those people a licensed adviser or primary sources are the better choice.
Download the skill