Improve package descriptions

scientific-packages/umap-learn/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: umap-learn
-description: Guide for using UMAP (Uniform Manifold Approximation and Projection) for dimensionality reduction, visualization, and clustering. Use this skill when working with high-dimensional data that needs to be reduced for visualization, machine learning pipelines, or clustering tasks. Triggers include requests for dimensionality reduction, manifold learning, data visualization in 2D/3D, UMAP-based clustering, or supervised feature engineering.
+description: Comprehensive guide for UMAP (Uniform Manifold Approximation and Projection) - a fast, scalable dimensionality reduction technique for visualization, clustering, and machine learning. Use this skill for: dimensionality reduction of high-dimensional datasets (genes, proteins, images, text embeddings, sensor data), creating 2D/3D visualizations of complex data, preprocessing data for clustering algorithms (especially HDBSCAN), supervised and semi-supervised dimensionality reduction with labels, transforming new data using trained UMAP models, parametric UMAP with neural networks, feature engineering for downstream ML models, manifold learning and non-linear dimensionality reduction, comparing UMAP to t-SNE/PCA/other methods, inverse transforms and data reconstruction, aligned UMAP for temporal/batch data analysis. Triggers include: "dimensionality reduction", "UMAP", "manifold learning", "data visualization", "clustering preprocessing", "high-dimensional data", "embedding", "reduce dimensions", "2D visualization", "3D visualization", "supervised dimensionality reduction", "parametric UMAP", "transform new data", "feature engineering", "HDBSCAN clustering", "t-SNE alternative", "non-linear dimensionality reduction", "inverse transform", "data reconstruction", "aligned embeddings", "batch effect correction", "temporal data analysis".
 ---
 
 # UMAP-Learn