Anthropic's official skill for building new Claude Code skills, testing them with evals and benchmarks, and tuning descriptions for reliable triggering.
Lets Claude drive a real browser to navigate sites, fill forms, capture screenshots and extract page data.
Analyzes CSV, Parquet and JSON data at maximum speed by intelligently routing between DuckDB and Polars — never pandas.
Search and rewrite code across 25 languages by syntax shape (AST) instead of text, with a safe two-pass codemod workflow.
Drive a real browser from Claude via the agent-browser CLI for navigation, form filling, screenshots, and data extraction.
Manus-style persistent planning that keeps task_plan.md, findings.md and progress.md on disk so work survives compaction and /clear.
End-to-end academic citation workflow: search papers, extract metadata, and generate validated BibTeX.
x
Generate and edit images across 30+ models (Gemini, Seedream, Recraft, GPT-Image, Riverflow) through one OpenRouter Image API call.
Query, download, and preview public NCI cancer imaging data (CT, MR, PET, digital pathology) using idc-index.
Convert PDFs, Office files, HTML, and more into clean Markdown with Microsoft MarkItDown for search, analysis, and LLM/RAG ingestion.
Search 11 scholarly APIs — PubMed, arXiv, OpenAlex, Semantic Scholar and more — for papers, citations, and open-access full text, with reproducible provenance.
An end-to-end skill for creating, editing, extracting from, and QA-ing PowerPoint .pptx/.potx files.
A guided statistics workflow that picks the right test, verifies assumptions, reports effect sizes, and writes APA-formatted results.
Plan studies with defensible sample sizes, minimum detectable effects, and power curves using both closed-form formulas and Monte Carlo simulation.
Turns vague questions into reproducible API queries across 78 documented public databases, complete with provenance.
A reference skill that guides classification, regression, clustering, model evaluation and pipeline building in scikit-learn with best practices baked in.
Runs bounded, network-free exploratory profiling of authorized local scientific files with missingness, leakage, and outlier diagnostics.
A three-phase Scan/Fix/Verify skill that finds WCAG 2.2 A and AA violations and writes framework-specific fixes for React, Vue, Angular, Svelte, Next.js and plain HTML.
Manus-style persistent planning that keeps task_plan.md, findings.md, and progress.md on disk so work survives /clear and context compaction.