Data analysis, machine learning, LLM apps, and prompt design (9.1K skills)
Turn blog data into inline SVG charts for HTML or MDX. Choose from seven chart types to show comparisons, trends, shares, or scores, with responsive sizing, screen reader descriptions, and source attribution.
Choose useful metrics for a startup or feature based on its business model and stage. The framework helps separate decision-making measures from vanity numbers, select one current priority metric with a guardrail, audit dashboards, and set baselines, targets, and instrumentation plans.
Calculate a controlled set of two-dimensional molecular descriptors and Morgan, RDKit, or MACCS fingerprints from standardized compounds. Review dataset quality through missing values, unusual ranges, repeated structures, and fingerprint density summaries.
Clean local chemical structure data from SMILES, CSV, SDF, or MolBlock files. Preserve each original record while standardizing structures, deriving parent forms, checking quality issues, and grouping duplicates for review or downstream analysis.
Search a standardized compound library for similar structures or substructures, group molecules, and select diverse representatives. The skill also reviews indexing and risk records, while keeping the library unchanged and recording the parameters and results.
Review structured chemical reaction records while preserving their original entries. Check participants, reported roles, yields, duplicates, and element or charge differences, then produce searchable records and a review queue with clear reasons.
Design, audit, troubleshoot, or refactor the software harness that coordinates an AI agent. It offers provider-neutral guidance for model loops, tools, permissions, memory, planning, connectors, observability, evaluation, and safety across different domains.
Prepare a vector geospatial dataset for a PortalJS portal by converting its coordinate system, creating a map-ready PMTiles file and a queryable GeoParquet file, and registering both with the original data for an automatically rendered showcase.
Generate DCAT catalog feeds for a PortalJS site so open data catalogs can harvest its datasets. The skill supports JSON-LD, Turtle, and RDF/XML formats, adds feed discovery, and checks conformance to selected profiles.
Create a new PortalJS data portal from a project brief. The skill gathers missing details about pages, datasets, language, and catalog organization, then applies them to the PortalJS template and checks the generated project with type and localization checks.
Register a CSV, TSV, JSON, or GeoJSON source in an existing PortalJS portal. The skill adds it to the dataset catalog and routes local files or remote URLs so the catalog and showcase can display the data.
Add a line, bar, area, pie, or scatter chart to the showcase for a dataset already registered in PortalJS. The skill creates a reusable Chart component and displays it in the dataset’s Views section using the existing data source.
Describe what a registered dataset contains by inferring field types and constraints from tabular data, then adding catalog details such as license, sources, and keywords to datasets.json. Choose a standard or custom profile so the PortalJS showcase can display typed fields.
Check a local or remote CSV or TSV for missing values, mixed types, duplicate rows, suspicious dates, and other common data issues. The skill returns a structured JSON report with column profiles and recommendations, while leaving the source file and project data unchanged.
Connect a PortalJS catalog to a live CKAN instance through its API. The skill generates a small server-side client and updates search and dataset showcase pages to fetch CKAN data, while pre-rendering pages for static deployment.
Plan a data portal before development by matching its purpose, data shape, audience, and constraints to choices for storage, catalog, compute, access, hosting, and metadata. The skill documents the recommendation and suggests the next build steps.
Add an interactive Leaflet map to a selected GeoJSON dataset’s PortalJS showcase. The skill can register a local or remote GeoJSON source, install the map dependencies, and render the map in the page’s Views section.
Migrate an ArcGIS Hub open data site into a PortalJS catalog. The skill collects the inventory, exports and converts spatial and tabular data for publishing, uploads files to Cloudflare R2 through Git LFS, and produces a report comparing source and derived data.
Add another file to an existing PortalJS dataset, such as a data dictionary, methodology document, or another period of data. The skill updates the dataset manifest and makes the showcase display a section for each resource.
Move datasets from CKAN, DCAT catalogs, Socrata, OpenDataSoft, or ArcGIS into a static PortalJS catalog or another CKAN instance. It previews the collected records, then writes dataset metadata and links or copied resource files to the chosen destination.
Create a domain-specific skill bundle through discussion and source-based research. The deliverable includes a concrete workflow and flowchart, an adapter, a trap fixture for testing common failures, and a smoke evaluation that compares results with and without the bundle.
Create and manage portable dstack presets for model inference optimization. A preset records the serving configuration, benchmark result, and hardware used, so teams can reuse the optimized setup across clouds, Kubernetes clusters, or bare-metal fleets.
Create agent skills with clear procedures and metadata that follow the agentskills.io specification. Use it to structure a new skill directory, write step-by-step instructions, and make a skill easier to find through its description.
Create or improve Claude skills with guidance on SKILL.md structure, reusable scripts, reference files, and packaging. The skill helps turn a task into practical agent instructions and resources, then refine them after use.