AI/ML

Transcribes video audio using WhisperX with accurate timestamp preservation. Creates JSON transcripts with word-level timing. Designed for parallel execution in video analysis pipelines where timestamp alignment is critical.

This skill implements a ReasoningBank system for AI agents to learn from task outcomes, recognize patterns, and optimize strategies. It provides methods for recording experiences, recommending approaches, and enabling continuous improvement through meta-learning and transfer learning capabilities.

Provides concrete patterns for deploying ML models to production, including data ingestion with dlt, batch/API deployment, drift detection, and incident response. Includes templates for common tasks and clear decision trees for choosing appropriate strategies. Covers both traditional ML and LLM/RAG security considerations.

This skill provides comprehensive access to PyTorch Geometric for graph neural network development. It covers installation, graph creation, dataset loading, model building, training loops, and benchmarking utilities. The documentation includes practical code examples for common GNN tasks like node classification and molecular property prediction.

A Python CLI tool that generates and edits images using OpenRouter's API with models like FLUX.2 Pro and Gemini 3 Pro. It handles both creation and modification tasks with clear model selection guidance and API key management.

Provides unified access to Google Gemini's multimodal capabilities for processing audio, images, videos, and documents, plus image generation. Handles files up to 9.5 hours audio, 6 hours video, 1000-page PDFs. Includes detailed model selection guides, code examples, and troubleshooting for common API issues.

This skill provides CLI access to Z.AI's multimodal capabilities including image/video analysis, OCR, web search, and GitHub exploration. It integrates with Claude Code via MCP, offering raw data output by default with JSON formatting option. Requires API key setup and Node.js environment.

This skill wraps the DiffDock tool for predicting 3D binding poses of small molecules to proteins using diffusion models. It handles PDB files or protein sequences and SMILES strings or ligand files, outputs multiple ranked poses with confidence scores, and supports batch processing for virtual screening. It clearly states it predicts poses, not binding affinity.

This skill enables differential gene expression analysis in Python using the PyDESeq2 library. It provides workflows for loading RNA-seq count data, specifying experimental designs, running statistical tests, and generating results with multiple testing correction. The documentation includes troubleshooting for common errors like design matrix issues and memory management tips.

This skill provides structured templates and specific techniques for writing effective prompts for GPT-5 and Claude. It focuses on clear instruction formatting, model-specific strategies like chain-of-thought, and iterative testing workflows to improve LLM outputs.

This skill provides structured guidance for deploying machine learning models to production. It covers model optimization, serving infrastructure setup, performance tuning, and monitoring. The checklist and workflow help engineers systematically address latency, throughput, and reliability requirements for real-world ML systems.

Generates Python practice exercises using evidence-based learning strategies like retrieval practice and spaced repetition. Follows an 'evals-first' design process, mapping exercises to learning objectives and success criteria. Provides structured templates for exercises and rubrics, supporting varied exercise types from fill-in-the-blank to AI-collaborative tasks.