wshobson quantitative-trading plugin

Version 1.2.3 of the plugin ships two skills and two subagents. backtesting-frameworks tabulates five backtesting biases with their mitigations — look-ahead against point-in-time data, survivorship against delisted securities, overfitting against out-of-sample testing, selection against pre-registration, transaction against realistic cost models — and diagrams walk-forward analysis over rolling train/test windows. risk-metrics-calculation groups metrics into volatility (standard deviation, beta), tail risk (VaR, CVaR), drawdown (max drawdown, Calmar) and risk-adjusted (Sharpe, Sortino), and warns against relying on VaR alone or assuming normal returns.

Each skill uses progressive disclosure: a 2–3.5 KB SKILL.md navigation tier over a references/details.md of about 18 KB (backtesting) and 17 KB (risk metrics) holding worked implementation patterns. The quant-analyst subagent specifies pandas/numpy/scipy, vectorized strategy implementation, backtests with transaction costs and slippage, out-of-sample testing and realistic assumptions about market microstructure; risk-manager covers R-multiples, position limits, hedging, expectancy calculation and stop-losses. Separate .claude-plugin/ and .codex-plugin/ manifests ship, and native plugin-install covers Codex, Cursor and Gemini as well as Claude Code. Install with /plugin marketplace add wshobson/agents, then /plugin install quantitative-trading.

The content is instructional prose with embedded Python examples — guidance for a model, not an importable library, with no tests or runnable package in the plugin. It is the layer for reasoning about a backtest rather than an engine that runs one, and the surface is small: two skills and two subagents. While the parent repository commits daily, the quantitative-trading subpath was last touched on 2026-05-29. The repository’s 38,858 stars belong to the whole 91-plugin claude-code-workflows marketplace, not to this plugin.