Guide · Agent skill
How to use the Investment Team Research agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The Investment Team Research skill (skill name investment-team) is an agent skill from the ai-berkshire repository that turns a request to research a public company into a team-style analysis. It is aimed at investors, analysts and students who want a structured first-pass research memo rather than a single loose answer. After this guide you will know what the skill does, how to install it, how to trigger it, and what it needs to run properly.
Investment Team Researchxbtlin/ai-berkshire · MIT Download the skill Quick start
- 1Install the skill. Download the ZIP from the skill page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps.
- 2Check the requirements. Confirm Python 3 is available, the ai-berkshire repository is checked out with its tools/ scripts, and the multi-agent tooling the workflow assumes is present.
- 3Enable web search for sub-agents. Add WebSearch to permissions.allow in .claude/settings.local.json so background sub-agents can search the web instead of silently failing.
- 4Ask for a company research run. Name the listed company you want researched in a plain request. The workflow shows the team structure and asks you to confirm.
- 5Wait for the four reports. The four sub-agents run in parallel and send their findings to the lead agent. The workflow shows progress as each report arrives.
- 6Review the final report. Read the conclusion, score table, bull and bear cases, checklist and recommendation, and check the stated data cutoff date and source notes.
- 7Run the data audit. Use report_audit.py to extract a random sample of figures, verify them against reliable sources, and get a pass or rework verdict before relying on the report.
What is an agent skill and what does this one do?
An agent skill is a folder containing a SKILL.md file that gives an AI agent a repeatable workflow to follow when a matching request comes in. This skill is generated from the source file skills/investment-team.md in the ai-berkshire repository, so Claude Code and Codex users share one workflow.
When you ask the agent to research a public company, the skill has it set up a team of four roles: business model and moat, financials and valuation, industry and competition, and risks and management. Four sub-agents run in parallel, each covering one area, and the main agent combines their findings into a long report. The report includes bull and bear cases, a checklist and an investment recommendation.
- Business analyst: business model and moat analysis, from a Duan Yongping perspective
- Financial analyst: financial statements and valuation, from a Buffett perspective
- Industry researcher: industry structure and competition, from a Munger perspective
- Risk assessor: risk and management assessment, from a Li Lu perspective
Install the investment-team skill
You can install the skill in one of two ways. Download the ZIP from the skill page on this site and unzip it into your personal skills folder at ~/.claude/skills/, or into a project's .claude/skills/ folder if you want it to apply only to that project.
Alternatively, in the Claude apps, open Settings, go to Capabilities, and upload the skill under Skills.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into the project's .claude/skills/ folder
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
Meet the requirements before you run it
The skill depends on more than the SKILL.md file. It runs scripts and uses the network, so check these requirements before you start a research run. The source workflow assumes multi-agent tooling named TeamCreate, Task and SendMessage.
The skill also needs access to the ai-berkshire repository files it references, including skills/financial-data.md and AGENTS.md, and the repository's tools/ scripts (financial_rigor.py, report_audit.py and twstock_data.py) run with Python 3.
- Python 3 and the repository's tools/ scripts: financial_rigor.py, report_audit.py, twstock_data.py
- Web search permission enabled for sub-agents
- Access to the ai-berkshire repository files: skills/financial-data.md and AGENTS.md
- Multi-agent or team tooling that the workflow assumes: TeamCreate, Task, SendMessage
Trigger the skill with a natural request
Name the company you want researched in a normal sentence. The skill's workflow takes the company name as the request. Before it starts, the workflow shows you the team structure and asks you to confirm.
Run the command from the repository root so the tool paths resolve. The skill's own guidance says to run the date command first so the report uses a confirmed data cutoff date.
- Research Meituan as a listed company using the investment team workflow
- Run a four-role investment team analysis on a public company I name
- Do a structured first-pass research memo on this stock with bull and bear cases
python3 tools/financial_rigor.py verify-market-cap --price {价格} --shares {股本} --reported {报告市值} --currency {币种}Check the web search permission before the run
Background sub-agents cannot ask you for permission interactively. If WebSearch is not in the permissions.allow list, their searches can be blocked silently. The skill's workflow therefore runs a pre-check first and stops if WebSearch is not allowed.
If the check fails, add "WebSearch" to permissions.allow in .claude/settings.local.json, or use /permissions to tick it, then run the request again. Without web search, the skill itself warns that sub-agents fall back to training knowledge and produce misleading reports.
grep -l '"WebSearch"' .claude/settings.local.json ~/.claude/settings.local.json 2>/dev/nullRead the output and the safety notes
The final report follows a fixed structure: a one-paragraph conclusion, a four-dimension score table, key financial data over two years, bull and bear points, a ten-item Buffett-style pre-purchase checklist, and a final recommendation with price ranges. The workflow also asks for a data audit step before publishing.
On safety, the skill runs local scripts and performs web searches, so review what it executes in your repository. The skill requires that financial data come from two independent sources, and that sub-agents must not present training knowledge as live search results. The report is AI-generated and may contain stale or unverified data.
- Check the data cutoff date stated in the report header
- Look for the confidence labels and source notes on key numbers
- Verify figures from the sources the report cites before you rely on them
What people use it for
- A first-pass structured memo on a listed company before deciding what to research further
- Studying how a value-investing framework applies to a real company, section by section
- Comparing the bull and bear arguments for a stock in one organized document
- Testing valuation arithmetic with the bundled tools instead of doing it by hand
Frequently asked questions
How do I install the investment-team agent skill?
Download the ZIP from the skill page and unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder for one project. In the Claude apps you can instead upload it under Settings → Capabilities → Skills.
Is the investment-team skill free to use?
The source material for this page does not state pricing or licence terms beyond the MIT licence listed for the repository. Check the repository's licence file for the terms that apply to you.
What do I need to run the investment-team skill?
You need Python 3, the repository's tools/ scripts including financial_rigor.py, report_audit.py and twstock_data.py, web search permission enabled for sub-agents, and access to skills/financial-data.md and AGENTS.md. The workflow also assumes TeamCreate, Task and SendMessage tooling.
Why do the sub-agents say they cannot search the web?
Background sub-agents cannot show permission prompts, so if WebSearch is not in permissions.allow they can be blocked silently. Add WebSearch to .claude/settings.local.json or use /permissions, then run the request again.
Can I use this skill to decide whether to buy or sell a stock?
No. The output is AI-generated and may contain stale or unverified data, so it should not replace professional advice. Use it as a structured first pass and verify key numbers yourself.
How is this different from asking a chatbot to analyse a stock?
The skill sets up four parallel sub-agents, each covering one area, then combines their findings into a fixed-structure report with bull and bear cases and a checklist. It also asks for numeric checks with bundled Python tools rather than mental arithmetic.
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, analysts and students who want a structured first-pass memo on a listed company and can run the repository's Python tools with web search enabled. It is a poor choice if you need a basis for real buy or sell decisions, if the company has thin public information, or if you cannot enable web search, since the output may then rely on outdated training knowledge.
Download the skill