Skill UI

This skill analyzes and reports on the token usage, expenditure, and budget for AI models (like Claude Code) by querying a local SQLite database. It is essential for users who need detailed financial accountability, tracking costs by project, tool, or date, and managing operational budgets.

Provides comprehensive configuration management for local LLM inference using RuVLLM. Users can select models, perform task-specific fine-tuning with MicroLoRA adapters, or implement continuous, real-time adaptation using SONA. This tool guides the entire advanced LLM workflow, covering status checks, configuration generation, and advanced adaptation loops.

Genomic Intelligence provides powerful, hosted transformer models for advanced genomic analysis. Users can input a DNA/FASTA sequence and receive structured predictions for key regulatory features, including promoter regions, splice sites, enhancer activity, chromatin state, gene expression (log TPM), and de-novo gene annotation. This service operates via a REST API or MCP server, eliminating the need for local GPU hardware or model weights, making complex bioinformatics accessible through sequence input.

This guide provides comprehensive methodological frameworks for conducting rigorous empirical analyses in international trade and macroeconomics, particularly tailored to the standards of top-tier journals like the JIE. It covers assembling bilateral trade and cross-country panels, estimating gravity models using PPML, implementing local projections, and calibrating structural open-economy models. It ensures analyses are robust, replicable, and address common referee criticisms.

A comprehensive guide detailing when and how to structure theoretical models in development economics research, particularly for World Bank Economic Review submissions. It emphasizes that models must serve a functional purpose—such as interpreting empirical coefficients, running policy counterfactuals, or aggregating to welfare statements—rather than being decorative. Key principles cover parameter identification, matching structural frictions to the local setting, and handling counterfactual assumptions (Lucas critique).

This tool de-identifies sensitive free-text columns (e.g., names, dates, identifiers) in local structured datasets (CSV, JSONL, Parquet) using advanced OpenMed models. It ensures strict privacy by generating a completely redacted dataset and a separate PHI-free aggregate summary, allowing for secure data sharing and analysis without compromising source privacy or logging sensitive values.

This skill de-identifies sensitive PII from clinical notes across numerous languages, including Spanish, German, French, and Arabic. It automatically selects language-specific PII models, national ID regex patterns, and generates locale-aware fake surrogates. Designed for global healthcare datasets, it runs entirely on-device, ensuring maximum privacy and linguistic accuracy for diverse international clinical records.

This skill manages the loading and efficient caching of OpenMed clinical/biomedical Named Entity Recognition (NER) models. Models are downloaded once from the Hugging Face Hub, enabling full offline operation afterward. It supports model reuse across multiple calls, configuring cache directories, specifying devices (CPU/GPU), and handling various model identifiers (registry keys, full HF IDs, or local paths).

Guides users to select a local, on-device Personally Identifiable Information (PII) detection model from a committed registry. It filters models based on language, required runtime environment (e.g., PyTorch, MLX), and size constraints. The process emphasizes building an offline shortlist and mandates rigorous pre-deployment recall validation and performance benchmarking to ensure secure, privacy-preserving deployment.

This skill manages complex, stateful proof projects as a local-first pipeline. It guides the user through drafting, rigorous verification, and source bookkeeping. It systematically handles proof continuation across runs, maintaining a full audit trail. When local attempts stall, it orchestrates handoffs to external AI models (like GPT Pro) or performs optional deep audits (DeepSeek), ensuring a structured, traceable, and verifiable output suitable for academic or technical submission.

This skill guides users on deploying OpenMed models completely on-device for maximum privacy and edge computing. It details running workflows using three backends: MLX (Apple Silicon), CoreML (iOS/macOS), and ONNX/WebGPU (cross-platform/browser). The process involves converting, optionally quantizing, and running models locally, ensuring PHI never leaves the host device.