Skill UI

This tool helps development teams identify gaps in their current test coverage. It analyzes code paths and detects areas that lack adequate unit or integration tests, generating actionable suggestions for improvement. Use it when assessing code quality after a feature implementation, or when a user needs to know which parts of the codebase require further testing.

This guide details the architectural and implementation patterns for building modular applications or microservices using the tinystruct Java framework. It covers core concepts like using the @Action annotation for dual routing (CLI/HTTP), managing request state with Context, and using native Builder for JSON serialization, ensuring a lightweight and structured development environment.

This skill intercepts the AI response flow to allow users to explicitly control the desired depth, length, and token usage of the generated answer before the model begins writing. By analyzing the prompt's complexity and the user's intent (e.g., 'short version', 'detailed answer'), it estimates the appropriate token range and prompts the user to select a granularity level (25%, 50%, 75%, or 100%).

This skill executes comprehensive historical backtesting using the high-performance neural-trader engine. It supports features like walk-forward validation and captures key financial metrics (Sharpe, Drawdown). Critically, it signs the resulting artifact using Ed25519 cryptography, providing non-repudiable tamper evidence required for promoting research models to live trading systems.

This skill dispatches computationally intensive quantitative trading jobs (such as multi-year walk-forward analysis, large Monte-Carlo simulations, parameter sweeping, or deep model training) to a managed cloud agent runtime. It is designed for large-scale, production-grade backtesting that exceeds local machine capacity, ensuring robust execution and artifact management via Anthropic's cloud services.

This skill calculates model interpretability for algorithmic trading signals using PageRank. It constructs a feature-contribution graph from an LSTM or Transformer signal and executes a single-entry PageRank calculation. The result identifies the top-K features that most significantly drove the prediction, providing a "regulator-grade" audit trail essential for compliance with regulations like the EU AI Act.

This tool automates the process of optimizing investment portfolio allocations using a mean-variance engine. It integrates neural prediction models and risk constraints to determine optimal asset weights. Users can set specific risk targets, assess comprehensive risk metrics (variance, correlation), and generate actionable, step-by-step rebalancing plans for professional portfolio management.

This skill performs mean-variance portfolio optimization by solving the linear system Σ · x = μ using the Conjugate Gradient (CG) method. It is significantly faster (40-60x) and numerically robust compared to legacy methods like the Neumann series. It takes a covariance matrix (Σ) and expected return vector (μ) to determine optimal asset weights, providing full provenance tracking for auditability.

A comprehensive tool for analyzing the current market regime (bull, bear, ranging, volatile) using advanced neural models. It detects the prevailing market structure, provides a confidence score, and recommends suitable trading strategies. Ideal for market analysis, backtesting preparation, or general market condition inquiries.

Assess and manage investment risk using the neural-trader risk engine. This tool calculates critical metrics such as Value at Risk (VaR), Conditional VaR (CVaR), and Sharpe ratio for both single positions and entire portfolios. It also optimizes position sizing using methods like Kelly criterion and checks for market circuit breaker triggers (e.g., high correlation spikes or volatility regime shifts) to ensure robust trading decisions and minimize exposure.

This tool generates sophisticated trading signals by leveraging an advanced anomaly detection engine. It scans financial assets for various market anomalies—such as spikes, drifts, and oscillations—and classifies them. It then integrates neural network prediction and searches historical pattern databases to output highly ranked, actionable trading signals, complete with confidence scores and target parameters.

A comprehensive tool for training and evaluating advanced neural network models (including LSTM, Transformer, and N-BEATS) on time-series market data. It facilitates the prediction of financial trends, generating results with confidence intervals, and comparing performance across multiple architectures. Ideal for quantitative trading research and financial forecasting.