skills
by ml4t
Agent skills for Machine Learning for Algorithmic Trading: leakage-safe quant ML, backtesting, validation, and autonomous research workflows.
47stars
13forks
Python
Added 11/15/2026
Available after the catalog security scan passes.
⚠️ Third-Party Software Notice
This skill is third-party open-source software developed and hosted independently on GitHub. SkillsLLM is an informational directory and does not control or maintain the underlying repository.
Any security checks, ratings, or warnings displayed by SkillsLLM are automated and limited in scope. They do not constitute a security certification or guarantee that the software is safe, error-free, or free from malicious code, vulnerabilities, compromised dependencies, or prompt-injection risks.
Review the source code, permissions, dependencies, and configuration before installing or running any third-party skill. Use is at your own risk. To the maximum extent permitted by applicable law, SkillsLLM is not liable for losses arising from third-party software.
The deep catalog scan for this skill is still queued. Run an instant dependency check now instead.
AI Agentsagent-skillsai-agentsalgorithmic-tradingbacktestingclaude-codecodexcoding-agentsfinancial-machine-learningmachine-learningml4tquantitative-finance
Maintainer of this skill?
Installation
# Add to your Claude Code skills
git clone https://github.com/ml4t/skillsGetting Started
Guides for using ai agents skills like skills.
- Caveman: Cut Claude Token Use by 65%How agent-side prompt compression works, when to use it, and when not to.
- How to Test an AI Skill: A Practical Guide to EvalsEvaluate skills with realistic fixtures, a baseline, and repeatable checks.
- Getting Started with AI SkillsFirst-time install walkthrough for Claude Code, Codex CLI, and ChatGPT.
Compare with similar skills
See how skills compares with popular alternatives.
ML4T Agent Skills
61 standalone skills that teach AI coding assistants the Machine Learning for Trading workflow from Machine Learning for Algorithmic Trading, 3rd Edition.
This repository turns the book's quant research discipline into runtime guidance for coding agents. Each skill gives an agent the pattern it needs to avoid common ML4T failures: leakage, lookahead bias, overfit backtests, stale data joins, broken cross-validation, unrealistic transaction costs, and unsafe autonomous workflows.
Each skill is a plain SKILL.md file with YAML frontmatter and concise procedural guidance. Installing them means putting each individual skill directory where your agent looks for skills, which scripts/install.sh does from a clone or from an unpacked release archive.
Next free session: How to use AI Agents for Trading Strategy Research, a 30-minute live session on Wednesday, October 28, 2026, 11:00 AM ET / 15:00 UTC. All courses, workshops, and free lessons.
Why Use These Skills
Coding agents are strong at writing code, but financial ML has failure modes that are easy to miss and expensive to discover late. These skills package the reviewer-grade checks that should happen before a strategy result is trusted:
- Method discipline: every skill teaches a concept-first WRONG/CORRECT pattern.
- Book alignment: skills map to the chapters and case-study workflow in Machine Learning for Algorithmic Trading.
- Production handoff: stable patterns point to the corresponding
ml4t-data,ml4t-engineer,ml4t-backtest,ml4t-diagnostic, orml4t-liveAPI. - Agent portability: the files are standard Markdown and can be used by Claude Code, Codex/OpenAI agents, or any tool that can discover and read local skill files.
Installation
The canonical distribution is this Git repository. There is no npm package, Python package, or build step, because the runtime artifact is the Markdown file itself.
Agents discover skills one directory level deep, at <skills-dir>/<skill-name>/SKILL.md. This repository groups skills into category directories so it can be read, so cloning it into a skills directory as a single folder will not work: the SKILL.md files end up three levels down and nothing finds them. scripts/install.sh flattens the categories for you.
git clone https://github.com/ml4t/skills.git ~/.ml4t-skills
cd ~/.ml4t-skills
./scripts/install.sh # ~/.claude/skills, the default
./scripts/install.sh .agents/skills # Codex, project-local
./scripts/install.sh ~/.claude/skills --copy # copies instead of symlinking
./scripts/install.sh --prune # also drop skills this repo no longer ships
./scripts/install.sh --uninstall # remove every skill it installed
The script symlinks each skill in as ml4t-<skill-name>, so git pull updates every installed skill at once, and the ml4t- prefix keeps them from colliding with skills you already have. It is idempotent and never overwrites a directory it did not create: a copy install carries a .ml4t-installed marker, and only a marked directory is ever replaced. Pass --copy when the install must not depend on the checkout.
A plain run only writes the skills being shipped now, so a skill that is renamed or dropped upstream stays installed until you say otherwise. --prune removes those leftovers as part of the install, and --uninstall removes the whole install. Both delete only what the script created, by the same symlink-or-marker test, and leave anything else in the directory alone.
Copying by hand works too. Move each of the 61 skill directories, not the category directories, into your agent's skills directory.
What these files can do
Nothing on their own. Every skill in this repository is Markdown: no executable scripts, no hooks, no MCP servers, no network calls. An empirical study of 31,132 public skills found that 26.1% contained at least one vulnerability, and that skills shipping executable scripts were 2.12 times more likely to be among them. Reading a skill before you install it is a reasonable habit, and here reading it is the whole audit.
How Skills Work
Every skill follows the same concept-first pattern:
- The Problem - what goes wrong without this knowledge
- WRONG code - the naive approach that looks right but fails
- CORRECT code - the right pattern using standard tools
- Guardrails - specific red flags to watch for
- Production Implementation - optional
ml4t-*library handoff when a stable API exists - Checklist - verification steps the agent can follow
The first 80% teaches the method with standard Python tools such as polars, numpy, scikit-learn, scipy, statsmodels, pytorch, and lightgbm. The final section, when present, points to the production-grade ml4t-* library implementation.
Skill Catalog
Concepts (10)
| Skill | What it prevents |
|---|---|
| backtest-overfitting | Selecting the best historical result without multiple-testing correction |
| causal-identification | Treating confounded associations as tradable causal effects |
| data-leakage | Train-test contamination via preprocessing, features, or labels |
| information-coefficient | Miscalculating or misinterpreting factor predictive power |
| lookahead-bias | Using future information in features, labels, or evaluation |
| non-stationarity | Ignoring changing statistical properties in financial time series |
| point-in-time | Using revised data that was not known at decision time |
| regime-awareness | Treating regimes as timing signals instead of conditioning variables |
| survivorship-bias | Excluding failed or delisted assets from historical analysis |
| transaction-costs | Backtesting signals that cannot survive spread, slippage, and impact |
Data (7)
| Skill | What it does |
|---|---|
| build-bars | Aggregates tick data into time, volume, and dollar bars |
| calendar-ops | Aligns data across trading calendars, holidays, and frequencies |
| continuous-futures | Builds roll-adjusted continuous futures series |
| data-export | Exports schema-validated datasets to efficient columnar storage |
| define-universe | Constructs point-in-time tradeable universes with liquidity filters |
| fetch-data | Fetches market data through provider abstractions with schema checks |
| validate-data | Checks gaps, stale prices, outliers, and OHLCV integrity |
Features (10)
| Skill | What it does |
|---|---|
| compute-features | Computes panel-aware financial features without cross-asset leakage |
| feature-families | Balances momentum, mean-reversion, volatility, carry, and value signals |
| feature-selection | Selects features inside CV folds using IC, MI, or RFE |
| feature-store | Stores versioned features with point-in-time retrieval |
| feature-validation | Audits IC significance, stability, redundancy, and leakage |
| horizon-design | Chooses prediction horizons from IC decay, turnover, and costs |
| latent-factors | Extracts latent factors with PCA, IPCA, or autoencoders |
| meta-labels | Uses a secondary model to filter or size primary trading signals |
| regime-features | Builds stationary volatility, trend, and liquidity regime features |
| triple-barrier | Labels trades with volatility-adaptive profit, stop, and time barriers |
Validation (8)
| Skill | What it does |
|---|---|
| cpcv | Runs Combinatorial Purged Cross-Validation |
| deflated-sharpe | Adjusts Sharpe ratios for multiple testing bias |
| drift-detection | Detects input, prediction, and concept drift |
| evaluate-factor | Evaluates factor IC, quantile spreads, turnover, and decay |
| purging-embargo | Removes overlapping labels and adds embargo buffers |
| shap-analysis | Explains model predictions with SHAP values |
| stationarity-tests | Applies ADF/KPSS tests and transformation decisions |
| walk-forward-cv | Evaluates time-series models with rolling or expanding windows |
Backtest (5)
| Skill | What it does |
|---|---|
| cost-model | Models commissions, slippage, and market impact |
| rl-execution | Applies reinforcement learning to execution and hedging |
| run-backtest | Runs event-driven backtests with Strategy/Broker/Engine contracts |
| sensitivity-analysis | Tests parameter robustness and overfitting cliffs |
| tearsheet | Generates performance reports from backtest returns |
Portfoli
Frequently Asked Questions
What is skills?
skills is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by ml4t. Agent skills for Machine Learning for Algorithmic Trading: leakage-safe quant ML, backtesting, validation, and autonomous research workflows. It has 47 GitHub stars.
Is skills safe to use?
skills's catalog security scan is still queued. You can run an instant dependency and prompt-injection check now with the "Scan for vulnerabilities" button above.
How do I install skills?
Clone the repository with "git clone https://github.com/ml4t/skills" and add it to your Claude Code skills directory (see the Installation section above).
What programming language is skills written in?
skills is primarily written in Python. It is open-source under ml4t on GitHub, so you can review or fork the full source.
Are there alternatives to skills?
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh skills against similar tools.
Comments (0)
to leave a comment.
No comments yet. Be the first to share your thoughts!
Related Skills
Reverse engineer anything with agents, from app behavior down to native binaries.
460,718
112,885
TypeScript
AI Agentsagent-skillsai
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
303,435
44,733
JavaScript
AI Agentsai-agentsanthropic
Skills for Real Engineers. Straight from my .agents directory.
250,518
21,171
Shell
AI Agents
An agentic skills framework & software development methodology that works.
234,966
20,863
Shell
AI Agentsaibrainstorming