Guide · Agent skill
How to use the Volatility Mean-Reversion Strategy agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The volatility skill gives an AI agent a simple trading-signal method based on mean reversion in historical volatility. It is written for quant researchers, developers and analysts who want a readable pandas example of how volatility regimes might relate to later price moves. This guide explains what the skill does, how to install it, how to ask for it, and where its limits are.
Volatility Mean-Reversion StrategyHKUDS/Vibe-Trading · MIT Download the skill Quick start
- 1Download the ZIP. Get the volatility skill ZIP file from the page where you found this guide.
- 2Install the skill. Unzip it into ~/.claude/skills/ for all projects, or into .claude/skills/ in one project. Alternatively, upload it in the Claude apps under Settings → Capabilities → Skills.
- 3Install Python packages. Run pip install pandas numpy if you do not already have them.
- 4Prepare OHLCV data. Supply your own price history with open, high, low, close and volume columns. The skill does not fetch data over the network.
- 5Ask for the signal. Request the volatility signal in plain language, naming your data and any settings such as annualize = 365 for crypto.
- 6Check and test. Read the code, confirm the warm-up zeros, and test the signals out of sample before relying on them.
What does the volatility skill make the agent do?
The skill builds a signal from price history using a volatility mean-reversion idea. It measures how volatile an asset has recently been as the annualized standard deviation of returns over a window of days (historical volatility, or HV). It then ranks that HV against the past lookback period as a percentile from 0 to 100.
Periods of unusually low volatility are marked as long, on the expectation that volatility will expand. Periods of unusually high volatility are marked as exit or short, on the expectation that volatility will contract. In the middle region the skill keeps the current position.
- Signal 1 means long (low-volatility regime).
- Signal -1 means short (high-volatility regime).
- Signal 0 means stand aside.
- The skill's own documentation says volatility is not direction: a long signal in a low-volatility regime does not guarantee price appreciation, only that volatility expansion is statistically more likely.
How do you install the volatility skill?
Download the ZIP file for the skill from the page where you found this guide. Then choose one of two routes.
For Claude Code, unzip the file into ~/.claude/skills/ to make it available to all of your projects, or into a project's .claude/skills/ folder to limit it to that project.
For the Claude apps, upload the skill under Settings → Capabilities → Skills instead.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into .claude/skills/ inside your project
- Claude apps: Settings → Capabilities → Skills, then upload
What does the skill need to run?
The skill needs Python with pandas and numpy. Install them with the command below if you do not already have them.
You also need OHLCV price data (open, high, low, close and volume) from your own source. The skill does not include market data. The skill's example engine file, example_signal_engine.py, computes the signals with pandas.
- Python with pandas and numpy
- OHLCV price data supplied by you
- The skill runs scripts, and does not use the network
pip install pandas numpyHow do you ask the agent to use the volatility skill?
Once the skill is installed, describe the job in plain language and name the data you have. The agent decides whether the skill applies based on its description, which covers volatility mean-reversion trading on any OHLCV data. Mentioning volatility percentiles or historical volatility makes a match more likely.
The default settings are a 20-day HV window, a 120-day lookback for ranking, a low threshold of 20 percentile, a high threshold of 80 percentile, and an annualization factor of 252. Use 365 for cryptocurrencies, because they trade every day of the week. The skill's source also notes that 252 is used for China A-shares.
- Example request: Using the volatility skill, compute HV signals from this OHLCV file with the default settings.
- Example request: Build volatility percentile signals for BTC daily closes with annualize set to 365.
- Example request: Explain which days this volatility signal would have gone long, short or flat over the last year.
- Example request: Change low_pct to 10 and high_pct to 90 and compare the signal counts.
What should you check before relying on the output?
Review any script before running it. The skill runs scripts, so read the code the agent produces and the example file before you execute it on your data. The skill does not use the network, so it will not fetch prices for you.
The signal is a simple rule, not a validated strategy. The skill has no risk management, position sizing, transaction costs or validation. Before you draw conclusions, test the signals out of sample and account for costs yourself.
Warm-up periods matter. Before the lookback window is filled there is not enough data to compute percentiles, so those signals should be 0 as the skill specifies. Do not count those early days as trades.
- Read scripts before running them.
- Treat the output as a hypothesis to test, not advice.
- Expect zeros during the warm-up period.
- Use annualize = 365 for crypto and check it for other markets.
What people use it for
- Prototyping a simple volatility-regime rule before writing a more complete backtest.
- Studying how low and high volatility percentiles relate to later price moves in a historical dataset.
- Learning the pandas steps for rolling standard deviation and rolling percentile rank from a readable example.
- Comparing how the annualization factor changes the HV values for equities versus cryptocurrencies.
Frequently asked questions
How do I install the volatility agent skill?
Download the ZIP from the page you found this guide on, then unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. In the Claude apps, you can instead upload it under Settings → Capabilities → Skills.
What do I need to run the volatility skill?
You need Python with pandas and numpy, which you can install with pip install pandas numpy. You also need your own OHLCV price data, since the skill does not supply any market data.
Is the volatility skill free and what licence does it use?
The source project is licensed under the MIT licence. This guide does not cover pricing for any other service, so check the page where you downloaded the skill for its terms.
Can I use the volatility skill for live trading?
The skill is not designed for live trading with real money. It gives a signal only, with no risk management, position sizing, costs or validation, and volatility does not guarantee the direction of price.
Why is my signal zero for the first days?
Percentiles need a full lookback window of history, so before that window is filled the signal is set to 0. This is expected behaviour described in the skill's pitfalls.
How is this different from other volatility strategies?
The skill is a simple, readable rule based on percentile ranking of historical volatility, with a short pandas example. It is a good starting point for learning, but it does not include the extra risk controls or validation tools that a production strategy would need.
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 are a quant researcher, developer or analyst who wants a readable, pandas-based volatility percentile signal to prototype and backtest. Pick something else if you plan to trade live with real money or you cannot evaluate backtests, because the skill is a simple rule without risk management, costs or validation.
Download the skill