lzwme/finance-quant-skills

What it solves

This repository provides a collection of specialized "Skills" for AI agents (specifically Claude) to bridge the gap between general-purpose LLMs and the technical requirements of quantitative financial trading. It enables AI agents to accurately interact with financial data APIs, use backtesting frameworks, and follow specific strategy development guidelines without requiring the user to manually provide extensive documentation for every session.

How it works

Built on the Agent Skills standard, the project organizes specific domain knowledge into folders containing SKILL.md instruction files. These skills can be installed into AI agents via tools like npx skills or as Claude Code plugins. Once installed, the agent dynamically loads the relevant instructions and resources to perform tasks such as querying stock data, writing trading code, or generating institutional-grade research reports.

Who it’s for

Quantitative traders, financial analysts, and developers who use AI agents to automate financial data analysis, develop trading strategies, and perform backtesting.

Highlights

• Comprehensive Data Integration: Includes skills for popular financial data interfaces like AKShare, BaoStock, Tushare, and jqdatasdk.
• Backtesting Support: Provides specialized instructions for using frameworks such as Backtrader and RQAlpha.
• Strategy Development: Includes guides and API references for QMT and JoinQuant strategy development.
• Multi-Agent Compatibility: Supports installation across different programming agents (e.g., Claude Code, Cursor) using symbolic links to maintain a single source of truth.