SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillsfinancial-data-analysis

financial-data-analysis

Data & analytics

View repository

Updated 10-15 09:17

About this skill

📊 Financial Time Series Analysis Toolbox

A comprehensive data analysis method toolbox covering stocks / commodity futures / cryptocurrencies / ETFs / forex / indices.

Data Acquisition Strategy

Asset Type Data Source Method
A-share quotes/financials/indices Tushare MCP tushare_daily, tushare_income and other MCP tools
A-share futures Tushare MCP tushare_fut_daily, tushare_fut_holding etc.
Hong Kong stocks/US stocks Tushare MCP tushare_hk_daily, tushare_us_daily
Macroeconomics Tushare MCP tushare_cn_gdp, tushare_shibor etc.
International commodity futures (WTI/gold, etc.) yfinance scripts/data_fetcher.py
Cryptocurrencies yfinance scripts/data_fetcher.py
Forex Tushare MCP or yfinance depending on the specific currency pair
Global indices Tushare MCP or yfinance tushare_index_global / yfinance

Rule: Before calling a tushare MCP tool, you must first ToolSearch("+tushare <关键词>") load. For the complete index of tushare tools, see stock-tushare-pro-mcp skill's reference/tool-index.md。

Analysis Method Routing

Based on user intent, read the corresponding references/methods/ document before performing the analysis:

User intent keywords Reference document Available scripts
Stationarity, trend tests, series decomposition, structural breaks, Hurst references/methods/01_time_series_fundamentals.md scripts/analysis_toolkit.py
Price forecasting, ARIMA, Prophet, VAR references/methods/02_forecasting.md scripts/analysis_toolkit.py
Correlation, cointegration, causality, lead-lag references/methods/03_cross_asset_relationships.md scripts/analysis_toolkit.py
Volatility, GARCH, VaR, tail risk references/methods/04_volatility_and_risk.md scripts/analysis_toolkit.py
Portfolio optimization, factor analysis, risk parity, efficient frontier references/methods/05_portfolio_and_factor.md scripts/analysis_toolkit.py
Market regime, regime, cycles, wavelets references/methods/06_regime_and_structure.md scripts/analysis_toolkit.py
Commodity seasonality, spreads, term structure, contango references/methods/07_commodity_specific.md scripts/analysis_toolkit.py
Network analysis, information flow, clustering, MST references/methods/08_network_and_information.md scripts/analysis_toolkit.py
Technical indicators (MA/RSI/MACD/KDJ/Bollinger Bands) references/methods/01_time_series_fundamentals.md scripts/indicators.py
Chart plotting, visualization references/visualization_cookbook.md —
Report format references/output_templates.md —

Execution Workflow

1. 识别用户意图 → 查上方路由表
2. 读取对应 references/methods/ 文档 → 选择合适方法
3. 获取数据：tushare MCP tool（优先）或 scripts/data_fetcher.py
4. 执行分析：scripts/analysis_toolkit.py 或 scripts/indicators.py
5. 生成图表：参照 references/visualization_cookbook.md
6. 输出报告：按 references/output_templates.md 格式

Constraints

MUST

  • Label data acquisition time and source (tushare / yfinance)
  • Attach a disclaimer to every report
  • Check series stationarity before analysis (when applicable)
  • Label and handle outliers

MUST NOT

  • ❌ Make guaranteed return promises
  • ❌ Fabricate or make up data
  • ❌ Ignore risk warnings
  • ❌ Guess key metrics when data is missing

Output Storage Specification

Output Directory

The default root directory is {output_dir}(configured by input-variables, default {workspace}/data/analysis/）。

{output_dir}/
├── reports/        # 分析报告 (.md)
├── charts/         # 图表文件 (.png)
├── datasets/       # 中间数据集 (.csv)
└── temp/           # 临时数据（可清理）

File Naming

{类型}_{标的}_{日期}.{格式}

Example:

  • report_CU_20260306.md
  • chart_AAPL_seasonal_20260306.png
  • dataset_corr_matrix_20260306.csv

Output Rules

Data Volume Handling Method
< 20 rows Display directly in the conversation, do not save to a file
>= 20 rows Save to datasets/, return the file path + summary
Charts Save to charts/, display embedded in the conversation
Analysis report Save to reports/, return the full report
Temporary/intermediate data Save to temp/, remind the user that it can be cleaned up

Collaborate with the tushare skill: Raw market data storage follows the tushare skill's output-storage.md specification ({workspace}/data/tushare/), this skill's {output_dir} only stores analysis results, not raw data, to avoid duplication.

Parameter usage

All configurable parameters are declared via input-variables , and AI obtains values in the following priority order during execution:

  1. Explicitly specified by the user in the conversation -> highest priority
  2. input-variables in default value -> fallback
# 用户说 "把分析结果存到 ~/Desktop/analysis"
-> output_dir = ~/Desktop/analysis

# 用户说 "分析铜价"
-> output_dir = {workspace}/data/analysis (使用默认值)
-> default_period = 1y (使用默认值)