VectorBT Backtesting Skills for Agentic Coding Tools

A comprehensive collection of backtesting skills for trading strategies using VectorBT. Works with 40+ AI coding agents via skills.sh — including Claude Code, Cursor, Codex, OpenCode, Cline, Windsurf, GitHub Copilot, Gemini CLI, Roo Code, and more.

Supports Indian markets, US markets, and Crypto markets with realistic transaction cost modeling, OpenAlgo indicators (100+, TA-Lib available on request), market-specific benchmarking, OpenStatz tearsheets, and robustness testing. Broker-neutral by design — fee models use industry-standard references that can be customized for any broker.

Quick Install

Install the skills into your project using npx skills. The CLI auto-detects your AI coding agent and installs skills to the correct directory.

# GitHub shorthand
npx skills add marketcalls/vectorbt-backtesting-skills

# Full GitHub URL
npx skills add https://github.com/marketcalls/vectorbt-backtesting-skills

Install a specific skill only:

npx skills add marketcalls/vectorbt-backtesting-skills -s backtest
npx skills add marketcalls/vectorbt-backtesting-skills -s optimize
npx skills add marketcalls/vectorbt-backtesting-skills -s vectorbt-expert
npx skills add marketcalls/vectorbt-backtesting-skills -s setup

List available skills before installing:

npx skills add marketcalls/vectorbt-backtesting-skills -l

Install globally (available across all projects):

npx skills add marketcalls/vectorbt-backtesting-skills -g

Supported AI Coding Agents

Skills are installed via skills.sh which supports 40+ agents. Each agent reads skills from its own directory:

| Agent | Skills Directory | |-------|-----------------| | Claude Code | .claude/skills/ | | Cursor | .agents/skills/ | | Codex | .agents/skills/ | | OpenCode | .agents/skills/ | | Cline | .agents/skills/ | | Windsurf | .agents/skills/ | | GitHub Copilot | .agents/skills/ | | Gemini CLI | .agents/skills/ | | Roo Code | .agents/skills/ | | + 30 more | Auto-detected by npx skills |

The npx skills add command detects which agents you have installed and places the skill files in the correct paths automatically.

Supported Markets

| Market | Data Source | Fee Reference | Default Benchmark | |--------|------------|---------------|-------------------| | India | OpenAlgo (NSE, BSE, NFO, MCX) | Delivery, Intraday, F&O (4-segment) | NIFTY 50 | | US | yfinance (NYSE, NASDAQ) | Stocks, Options, Futures (per-share + per-contract) | S&P 500 | | Crypto | yfinance / CCXT | Spot, Perpetual Futures (maker/taker) | Bitcoin | | Custom | Any provider via extensible pattern | User-defined | User-defined |

Broker-neutral: Fee models use realistic industry-standard defaults and can be adjusted for any broker by changing the fees and fixed_fees constants.

Capabilities

Skills (User-Invocable Commands)

| Command | What It Does | |---------|-------------| | /setup | Detects OS, creates venv, installs OpenAlgo + all packages (TA-Lib optional), creates backtesting/ folders, configures .env with API keys | | /backtest | Generates a complete backtest script with signals, market-specific fees, benchmark comparison, plain-language report, OpenStatz tearsheet | | /optimize | Parameter grid search with OpenAlgo ta indicators, tqdm progress bars, Plotly heatmaps, best params vs benchmark | | /quick-stats | Inline code block (no file) — fetch data, run EMA crossover, print compact stats + benchmark alpha | | /strategy-compare | Side-by-side comparison of multiple strategies on same symbol, overlaid equity curves |

Pre-Built Strategy Templates (12)

| Strategy | Type | Description | |----------|------|-------------| | EMA Crossover | Trend | EMA 10/20 crossover | | RSI | Mean-reversion | RSI(14) oversold/overbought | | Donchian Channel | Breakout | Channel breakout with shifted levels (no lookahead) | | Supertrend | Trend | Supertrend with intraday session windows (9:30-15:00, exit 15:15) | | MACD | Trend + Breakout | MACD zero-line regime + signal-candle breakout | | SDA2 | Trend | WMA + STDDEV + ATR band system | | Double Momentum | Momentum | MOM + MOM-of-MOM with next-bar fill | | Dual Momentum | Rotation | Quarterly ETF rotation (NIFTYBEES vs GOLDBEES) | | Buy & Hold | Passive | Static multi-asset allocation with FD benchmark | | RSI Accumulation | Accumulation | Weekly RSI slab-wise buying (5%/10%/20% by RSI level) | | Walk-Forward | Validation | Rolling train/test optimization with WFE scoring | | Realistic Costs | Analysis | Same strategy across 5 fee tiers (zero to full delivery) |

Knowledge Base (20 Rule Files)

| Category | What's Covered | |----------|---------------| | Data | OpenAlgo (India), yfinance (US/Global), CCXT (Crypto), custom providers, .env + python-dotenv, CSV loading, resampling | | Indicators | OpenAlgo ta by default - 100+ indicators (EMA, SMA, RSI, MACD, BBands, ATR, ADX, STDDEV, MOM, Supertrend, Donchian, Ichimoku, HMA, KAMA, ALMA, ZLEMA, VWMA, oscillators, statistical, hybrid). TA-Lib only if explicitly requested | | Signals | ta.exrem() signal cleaning, ta.crossover(), ta.crossunder(), ta.flip() regime detection | | Simulation | from_signals, from_orders, from_holding, long/short/both directions | | Sizing | Percent, Value, TargetPercent, whole shares (min_size=1), futures lot sizes, fractional crypto | | Costs | India: 4-segment model (Delivery 0.111%, Intraday 0.0225%, Futures 0.018%, Options 0.098%). US: Per-share + per-contract model (Stocks ~0.01%, Options ~0.2%, Futures ~0.001%). Crypto: Maker/taker model (Spot 0.1%, Futures 0.02%/0.05%, funding rates). All customizable. | | Futures | SEBI revised lot sizes (Dec 2025): NIFTY=65, BANKNIFTY=30, FINNIFTY=60. US: E-mini/Micro contract specs | | Risk | Stop loss, take profit, trailing stop (sl_trail) | | Optimization | Loop-based (OpenAlgo ta / TA-Lib compliant) + broadcasting (vbt.MA exception for parameter sweeps) | | Benchmarking | India: NIFTY 50 via OpenAlgo. US: S&P 500 (^GSPC). Crypto: Bitcoin (BTC-USD). Strategy vs Benchmark table always produced | | Reporting | Plain-language backtest explanation for normal traders. OpenStatz HTML tearsheets with 30+ metrics, Monte Carlo simulations | | Plotting | Plotly dark theme, candlestick with xaxis type="category" (no weekend gaps), VectorBT 7-panel plot pack | | Validation | Walk-forward analysis (WFE ratio), robustness testing (Monte Carlo trade shuffle, noise injection, parameter sensitivity, entry/exit delay, cross-symbol validation) | | Safety | 10 common pitfalls with prevention, checklist before going live |

Prerequisites

1. AI Coding Agent

Install any supported AI coding agent. For example:

• Claude Code — npm install -g @anthropic-ai/claude-code
• Cursor — Desktop IDE with built-in AI
• Codex — npm install -g @openai/codex
• OpenCode — go install github.com/opencode-ai/opencode@latest
• Cline — VS Code extension
• Windsurf — Desktop IDE with AI
• Or any of the 40+ supported agents

Then install the skills:

npx skills add marketcalls/vectorbt-backtesting-skills

2. Data Source Setup

Indian Markets — requires OpenAlgo:

git clone https://github.com/marketcalls/openalgo.git
cd openalgo
pip install -r requirements.txt
python app.py

OpenAlgo runs locally at http://127.0.0.1:5000. You need a broker account connected via OpenAlgo and an API key from the dashboard. See OpenAlgo documentation.

US Markets — no setup needed. Uses yfinance (public Yahoo Finance data).

Crypto Markets — no setup needed for public data (yfinance or CCXT). Exchange API keys are optional (only for private endpoints).

3. Python Environment Setup

Use the /setup skill for automated setup, or manually:

python -m venv venv
source venv/bin/activate   # Linux/Mac
# venv\Scripts\activate    # Windows

# Install Python packages (OpenAlgo ta is the default indicator library - no C library needed)
pip install openalgo vectorbt plotly anywidget nbformat pandas numpy yfinance python-dotenv tqdm scipy numba ipywidgets openstatz ccxt

# Optional: only if you want to be able to request TA-Lib explicitly in a backtest
brew install ta-lib         # macOS
# sudo apt install libta-lib-dev  # Linux
pip install ta-lib

4. Configure API Keys

cp .env.sample .env
# Edit .env with your API keys

Usage Examples

/setup - Environment Setup

Detects OS, creates venv, installs dependencies, creates folder structure, and collects API keys into .env.

/setup
/setup python3.12

/backtest - Quick Backtest

Create a complete backtest script with market-specific fees, benchmark comparison, plain-language report, and OpenStatz tearsheet.

# Indian Markets
/backtest ema-crossover SBIN NSE D
/backtest rsi RELIANCE NSE D
/backtest supertrend NIFTY NFO 5m

# US Markets
/backtest ema-crossover AAPL
/backtest rsi MSFT

# Crypto Markets
/backtest ema-crossover BTC-USD

/optimize - Parameter Optimization

Optimize strategy parameters, generate Plotly heatmaps, and compare best parameters vs benchmark.

/optimize ema-crossover SBIN NSE D
/optimize rsi AAPL

/quick-stats - Inline Stats

Print key backtest stats with benchmark comparison without creating a file.

/quick-stats RELIANCE
/quick-stats AAPL
/quick-stats BTC-USD

`/strategy-comp