Skill UI

A comprehensive framework designed to structure and automate PyTorch deep learning workflows. It simplifies boilerplate code by organizing models into LightningModules, automating training loops via the Trainer, and providing robust support for multi-GPU/TPU scaling (DDP, FSDP, DeepSpeed), data handling (DataModules), and best-in-class experiment logging (W&B, MLflow). Ideal for professional, scalable deep learning project development.

Pyzotero is a powerful Python wrapper for the Zotero Web API v3. It enables developers to programmatically interact with Zotero libraries, allowing for CRUD operations on items, collections, tags, and attachments. Use this skill to build robust research automation workflows, manage bibliographic references, export citations (e.g., BibTeX), or perform bulk data processing within a Zotero environment.

This specialized skill automates the process of converting visual designs (from Stitch) into high-quality, modular React and Vite components. It adheres to modern frontend best practices, enforcing logic isolation via custom hooks, data decoupling using mock data files, and ensuring type safety with TypeScript interfaces. The workflow includes automated asset fetching, style mapping (Tailwind), and rigorous validation cycles to produce production-ready, maintainable code.

A powerful AI tool designed to transform complex scientific concepts into publication-quality visual schematics. Utilizing Nano Banana 2 AI for generation and Gemini 3.1 Pro Preview for rigorous quality review, the skill specializes in creating architectural diagrams, neural network structures, biological pathways, and complex flowcharts. It employs smart iterative refinement, ensuring the output meets predefined quality thresholds for academic journals, posters, and reports.

This core skill is designed for generating complete, publication-ready scientific manuscripts. It guides users through a rigorous two-stage process: first, creating structured outlines using deep research lookup, and second, converting these into flowing, authoritative prose. It supports IMRAD structure, advanced citation formatting (APA, AMA, Vancouver), and mandatorily integrates complex visual elements like graphical abstracts and technical schematics, ensuring the final output meets high journal submission standards.

UMAP (Uniform Manifold Approximation and Projection) is a powerful technique for non-linear dimensionality reduction. This skill enables the creation of scalable, structure-preserving embeddings (2D/3D) for visualization, clustering preprocessing, and advanced machine learning feature engineering. It supports various metrics (e.g., Euclidean, Cosine) and supervised learning workflows.

This advanced skill optimizes LLM context usage by implementing strict rules for prompt caching, efficient log pruning, AST-based code inspection, and compact data structuring. It minimizes token consumption and maximizes information density, suitable for highly technical and iterative development workflows.

Access a comprehensive network of 78 public databases across scientific, biomedical, environmental, and physical domains. This skill enables deterministic and reproducible data retrieval through documented REST APIs. It manages complex workflows including exhaustive pagination, cross-referencing identifiers across sources (e.g., PubChem, UniProt), and generating full provenance logs, ensuring results are auditable for high-stakes scientific research.

This tool automates the complete specification generation process across four core phases: Initialization, Requirements, Design, and Tasks. It supports both interactive (step-by-step approval) and automatic execution modes. After generating the full spec, it performs a final sanity review to ensure coherence and completeness, providing a ready-to-implement blueprint.

This skill automates the process of transforming high-level product specifications, requirements, and design documents into a detailed, actionable, and executable task plan. It follows a multi-stage workflow, including context loading, parallel research, and two rigorous review gates (Task Plan Review and Task-Graph Sanity Review). The output provides properly sized (1-3 hours), logically sequenced tasks, ensuring every requirement is covered and dependencies are explicitly managed, dramatically improving project planning accuracy.

This expert acts as a Design Systems Lead and Prompt Engineer for creating high-fidelity, consistent UI designs. It transforms vague ideas into structured specifications by enhancing prompts with professional UI/UX terminology, synthesizing design system source files, and managing multi-step generation and editing workflows using the Stitch MCP.

This guide details setting up a robust, local development loop for Adobe integrations (Firefly, PDF, Photoshop). It covers using the `aio` CLI for App Builder, implementing hot reload, and establishing unit/integration testing with mocking to ensure rapid, reliable iteration cycles for Adobe APIs.