Guide · Agent skill
How to use the Asset Allocation and Optimizers agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The asset-allocation skill is a reference guide that an AI agent can read to explain classic portfolio construction methods and help you configure one of five built-in optimizers. It covers Modern Portfolio Theory, Black-Litterman, risk budgeting, the all-weather approach, and the parameters and defaults for each optimizer. By the end of this guide you will know what the skill does, how to install it, how to ask for it in plain language, and what it does not do.
Asset Allocation and OptimizersHKUDS/Vibe-Trading · MIT Download the skill Quick start
- 1Download the ZIP. Download the skill ZIP from this page.
- 2Unzip into a skills folder. Extract the asset-allocation folder into ~/.claude/skills/ for personal use, or into .claude/skills/ inside a project.
- 3Or upload in the Claude apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the skill.
- 4Ask a specific question. Name the framework or optimizer you want, such as risk parity or turnover_aware, and ask for an explanation or a config setting.
- 5Write settings into config.json. Set optimizer and optimizer_params in config.json, starting from the defaults listed in the skill.
- 6Backtest before relying on it. Test the chosen optimizer on your own data, since the skill warns that mean-variance is often poor out of sample.
What the asset-allocation skill makes the agent do
An agent skill is a folder containing a SKILL.md file with instructions that the agent reads when a task matches the skill's topic. This skill's SKILL.md covers asset allocation theory and the usage of the optimizers built into the Vibe-Trading system.
The skill explains four classical frameworks: Modern Portfolio Theory (MPT, Markowitz), Black-Litterman, risk budgeting, and the all-weather strategy. It also gives a guide to five optimizers: equal_volatility, risk_parity, mean_variance, max_diversification and turnover_aware. The output is meant to be written into a config.json file.
- MPT: maximize expected return for a given level of risk, with the efficient frontier as the main idea.
- Black-Litterman: start from market equilibrium returns and blend in your own views.
- Risk budgeting: allocate by risk contribution rather than by capital share.
- All-weather: allocate risk equally across economic environments.
- Five optimizers: each with its principle, parameters, default values and trade-offs.
Install the asset-allocation skill in Claude
Download the ZIP file for the skill from this page. Unzip it so that the asset-allocation folder containing SKILL.md sits in one of two places: your personal skills folder at ~/.claude/skills/, or a project-specific folder at .claude/skills/ inside your project.
If you use the Claude apps instead, you can upload the skill under Settings → Capabilities → Skills rather than placing files on disk.
- Personal install: put the folder in ~/.claude/skills/ so it is available across your projects.
- Project install: put the folder in .claude/skills/ inside the project so it is available only there.
- Claude apps: upload the skill under Settings → Capabilities → Skills.
How to trigger the skill with natural requests
You do not need a special command. Ask about the topics the skill covers, and the agent can use its instructions when they match your request. Naming the framework or optimizer you care about makes the match more likely.
Good requests are specific about what you want: an explanation, a comparison, or a config setting. The skill's output is designed to be written into config.json, so asking for the exact settings is useful.
- Explain the difference between risk parity and mean-variance in plain language.
- Which optimizer should I use if I have no return forecasts but care about correlation?
- Write an optimizer_params block for turnover_aware with a lookback of 60 and a turnover_penalty suited to daily data.
- What does Black-Litterman's tau parameter do, and what range is typical?
Choose and configure one of the five optimizers
Optimizers are set in config.json through two keys: optimizer, which names the method, and optimizer_params, which holds its settings. The skill lists these defaults: lookback is 60 for every optimizer, risk_free is 0.0 for mean_variance, and risk_aversion is 1.0 and turnover_penalty is 0.0 for turnover_aware.
The skill's decision tree starts from whether you have return forecasts. If you do not, it asks whether you need to account for correlation; if so, it names risk_parity as the recommended default.
- equal_volatility: weights proportional to inverse volatility; no return forecast or correlation matrix needed.
- risk_parity: each asset contributes the same risk; accounts for correlation but needs iterative solving.
- mean_variance: maximizes the Sharpe ratio; sensitive to inputs and often performs poorly out of sample.
- max_diversification: maximizes the diversification ratio; limited effectiveness when assets are highly correlated.
- turnover_aware: mean-variance utility with a penalty on weight changes versus the previous rebalance; tune turnover_penalty to your data frequency.
{"optimizer": "equal_volatility", "optimizer_params": {"lookback": 60}}Understand the theory before you change settings
The skill warns that raw MPT often produces extreme weights and is extremely sensitive to inputs. Its practical advice is to add constraints such as upper and lower bounds or sector limits, or to use a regularized version.
For mean_variance, the skill recommends not making lookback too short, because a window under 30 easily overfits, and adding upper and lower weight constraints. For turnover_aware, it notes that with daily returns even a turnover_penalty near 0.5 strongly prefers holding still.
- Do not run raw MPT without constraints.
- Avoid very short lookback windows for mean_variance.
- Re-tune turnover_penalty whenever the return frequency of your data changes.
Safety notes: scripts, network and limits
According to the skill's listing, it runs no scripts and uses no network. The SKILL.md file itself contains no executable code, and the only file in the package is SKILL.md.
The skill gives general theory and example allocations. It is not personal financial advice and does not account for your circumstances or risk tolerance. Its own guidance says mean-variance is often poor out of sample and sensitive to inputs, so test any settings with backtesting before relying on them.
- No scripts run and no network access is used, according to the listing.
- Settings only take effect inside the Vibe-Trading system.
- Example allocations such as the China-focused all-weather mix are illustrations, not recommendations.
What people use it for
- A student comparing MPT, Black-Litterman and risk parity for a coursework essay.
- A developer choosing between risk_parity and turnover_aware for the Vibe-Trading optimizer settings.
- An investor asking for a plain-language explanation of risk parity versus mean-variance before reading further.
- An analyst drafting an optimizer_params block to paste into config.json.
Frequently asked questions
How do I install the asset-allocation skill?
Download the ZIP from this page and unzip the asset-allocation folder into ~/.claude/skills/ for personal use, or into .claude/skills/ inside a project. In the Claude apps you can instead upload it under Settings → Capabilities → Skills.
What does the asset-allocation skill need to run?
The skill requires the Vibe-Trading system for its config.json optimizer settings to have effect. The listing says it runs no scripts and uses no network.
Is the asset-allocation skill free?
The skill is published under the MIT licence. The source material does not mention pricing, so check the project page for any terms beyond the licence.
Which optimizer should I pick?
The skill's decision tree suggests risk_parity as the recommended default when you have no return forecasts but need to account for correlation. If you have return forecasts and trading costs matter, it points to turnover_aware.
Can the skill give me personal investment advice?
No. The skill gives general theory and example allocations. It does not account for individual circumstances or risk tolerance, so it is not a substitute for personal financial advice.
How does this compare with other allocation tools?
The source does not compare this skill with other tools. It is best understood as a reference guide for the optimizers built into Vibe-Trading, rather than a standalone calculator.
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 students, analysts and developers who want clear explanations of portfolio frameworks and help writing optimizer settings for the Vibe-Trading system. It is not suited to anyone wanting tailored investment advice or ready-to-trade decisions, and any settings should be backtested on your own data first.
Download the skill