Guide · Agent skill
How to use the Fund Analysis and Screening agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The fund-analysis skill is a set of written instructions that teaches an AI agent a structured method for judging mutual funds, ETFs and fund-of-funds (FOF) portfolios, mainly for China's A-share market. It is aimed at individual investors, analysts and advisers who want a repeatable checklist. After this guide you will know what the skill does, how to install it, how to phrase requests so the agent uses it, and what to check before relying on its output.
Fund Analysis and ScreeningHKUDS/Vibe-Trading · MIT Download the skill Quick start
- 1Download the ZIP. Get the fund-analysis skill ZIP from this page.
- 2Unzip into a skills folder. Place the fund-analysis folder in ~/.claude/skills/ or in your project's .claude/skills/ directory.
- 3Or upload in the Claude apps. Open Settings, go to Capabilities, then Skills, and upload the skill.
- 4Install Python libraries. Run pip install pandas numpy scipy so the skill's scripts can run.
- 5Provide your data. Give the agent fund return, benchmark and holdings data, since the skill does not supply live figures.
- 6Ask a specific question. Name the fund or ETF, the period and the decision, then review the report against the source data.
What is an agent skill, and what does fund-analysis make the agent do?
An agent skill is a folder containing a SKILL.md file that tells an AI agent how to handle a certain kind of task. When the skill is installed and relevant, the agent follows its framework instead of improvising. The fund-analysis skill is categorised as asset-class and its stated goal is to find sustainable excess returns rather than simply picking the funds with the best past performance.
Once installed, the agent can calculate and explain metrics such as annualised return, alpha, information ratio, max drawdown, Sharpe ratio, Sortino ratio and Calmar ratio. It can run a style-box analysis to check whether a fund's actual holdings match its label, watch for style drift over rolling windows, review a fund manager's track record, screen funds through a five-step framework, assess ETF tracking efficiency and suggest a fund-of-funds allocation with rebalancing rules.
- Performance metrics with suggested threshold values for equity funds
- Sharpe-style box regression to classify a fund as large or small cap and value or growth
- Style drift detection using a 60-trading-day rolling window with a 20-day step
- Fund manager evaluation including alpha, capture ratios and turnover
- ETF selection criteria covering tracking error, fees, liquidity and fund size
- FOF allocation templates and rebalancing triggers
Install the fund-analysis skill
Download the ZIP of the fund-analysis skill from this page. Then choose one of the two install routes described below.
For a personal setup, unzip the file into ~/.claude/skills/ so the fund-analysis folder sits inside that directory. For a single project, unzip it into the project's .claude/skills/ folder instead. 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 .claude/skills/ inside your project
- Claude apps: Settings → Capabilities → Skills, then upload
Install the Python dependencies
The skill runs scripts, and it needs Python with three libraries. Install them with pip before your first analysis. The skill does not use the network, so the libraries are the only installation step beyond the ZIP.
pip install pandas numpy scipySupply the data the analysis needs
The skill contains methodology and example data only. It does not give the agent live fund figures. You need to provide fund return, benchmark and holdings data yourself, or point the agent to a data source it can access. Without this input the agent can explain the framework but cannot produce real metrics for a specific fund.
Expect the example fund data in the skill to be possibly outdated. Use it to understand the output format, not as current facts about a product.
- Fund return history for the fund and its benchmark
- Holdings data if you want style-box and concentration checks
- Fee and fund size information for cost and liquidity checks
How to trigger the skill with natural requests
You do not need a special command. Ask for a fund or ETF analysis in plain language that names the kind of task, and the agent should pick up the skill from its description. Name the fund or ETF code, the period you care about and the decision you are making.
Useful request patterns include asking for a full report on one fund, asking whether a fund's style matches its label, asking the agent to screen a list, asking for an ETF comparison, or asking for a fund-of-funds allocation for a given risk level.
- Analyse fund 005827 over the last three years and give me a report with Sharpe ratio and max drawdown.
- Check whether this large-value fund has drifted toward growth over the past year.
- Screen these equity funds using the five-step framework and list the ones that pass each step.
- Compare the three CSI 300 ETFs by tracking error, fees and liquidity.
- Build a balanced fund-of-funds allocation with quarterly rebalancing rules.
Read the output and check it
The skill defines a report layout with sections for fund overview, performance over three years, style analysis and an evaluation with strengths, weaknesses and a suggestion. Treat each figure as a result of the inputs you supplied and the formulas in the skill. Check the period, benchmark and data source before acting on it.
The thresholds and allocation percentages in the skill are illustrative rules of thumb, not guarantees. The skill also lists pitfalls to check against, including survivorship bias, the effect of fund size above 200 billion yuan, quarter-end window dressing, subscription and redemption impact, fee drag over long periods and index-enhanced funds that deviate far from their benchmark.
- Confirm the benchmark matches the fund type (for example CSI 300 for equity funds)
- Cross-check quarter-end figures with mid-month data
- Watch for fees and tracking error that conflict with the skill's thresholds
What people use it for
- Screening a shortlist of A-share equity funds with the five-step framework before choosing one
- Checking whether a fund labelled large-cap value has actually drifted toward growth holdings
- Comparing broad-market ETFs on tracking error, management fee, daily trading value and fund size
- Drafting a balanced fund-of-funds allocation with quarterly drift checks and a 5% rebalancing trigger
- Teaching students how Sharpe ratio, information ratio and style-box regression are calculated and interpreted
Frequently asked questions
Is the fund-analysis skill free to use?
The skill is published under the MIT licence, as stated on this page. The skill itself does not require payment, but you still need your own fund data and a Python environment with pandas, numpy and scipy.
What do I need to run the fund-analysis skill?
You need Python with pandas, numpy and scipy, installed with pip install pandas numpy scipy. You also need fund return, benchmark and holdings data supplied by you or a source the agent can access, because the skill contains only methodology and example data.
Does fund-analysis fetch live fund data from the internet?
No. The skill does not use the network, and its file contains only methodology and examples. The agent must obtain actual figures from elsewhere before it can analyse a specific fund.
Can I use this skill for US or European funds?
The skill is designed for China A-share products. Its benchmarks, ETF codes and fee examples are China-specific, so you would need to adapt the benchmarks and data before using it for other markets.
How do I install an agent skill like this one?
Download the ZIP and unzip it into ~/.claude/skills/ for personal use or into .claude/skills/ inside a project. In the Claude apps you can instead go to Settings, Capabilities, Skills and upload it.
Why does the agent not return the same numbers as my fund platform?
The figures depend on the data you supply and the exact period, benchmark and formulas used. The skill's example fund data may be outdated, and differences in return or fee data sources can change results.
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 fund-analysis if you want a structured, repeatable checklist for comparing A-share mutual funds and ETFs, or a starting template for a fund-of-funds process, and you are comfortable supplying your own data. Pick something else if you need live market data, non-Chinese markets out of the box, or a ready-made investment decision, since the thresholds and allocations are rules of thumb and the example figures may be stale.
Download the skill