Skill UI

This skill provides a structured framework for designing and implementing composable recommendation, ranking, and feed pipelines. It models the industry-standard six-stage pattern: Source, Hydrator, Filter, Scorer, Selector, and SideEffect. Use it for any system that selects "the top K items for a (user, context)", such as social feeds, ad ranking, RAG rerankers, or content recommendation engines, ensuring a robust and scalable architecture.

This skill solves the mean-variance portfolio optimization problem (Σ · x = μ) using the highly efficient Conjugate Gradient (CG) method. It significantly outperforms legacy Neumann series methods, providing state-of-the-art speedup (40-60x) and accuracy for calculating optimal investment weights. It is designed for quantitative finance and risk management applications where computational performance is critical.

Provides a structured workflow for systematically accessing and analyzing official intellectual property data from the USPTO. This skill covers patent search (PatentSearch API), trademark status checks (TSDR), ownership transfer tracking, and building reproducible IP landscape logs. It is designed for data collection and record verification, emphasizing the use of official APIs and rigorous logging practices. (Warning: Not for legal advice.)

This skill clusters a collection of vectors (embeddings, e.g., code snippets) within a designated namespace using graph community detection algorithms (Spectral/Louvain). It is essential for discovering underlying themes, identifying natural groupings, and analyzing the relationships within large, unstructured vector datasets. Use it when you need to organize embeddings based on semantic relatedness rather than simple keyword matching.