Turning data into a visual language for better thinking.
In the era of data-driven decision-making, efficient and accurate data visualization and analysis are paramount. AntV offers a professional suite of visualization solutions, providing a robust toolkit and a comprehensive set of skills for the entire workflow—from chart design and interactive exploration to in-depth data analysis. It empowers users to swiftly transform complex datasets into intuitive visual charts, significantly lowering the barrier to creation through intelligent design specifications and a rich library of components. Whether for daily reporting, dynamic dashboards, or sophisticated interactive analysis, AntV delivers reliable support. By integrating AI capabilities, these tools further streamline and automate the generation and optimization of visualizations. This allows analysts to focus more on uncovering insights and driving business decisions, truly making data visible and understandable.
Warning
This project only merges AI-generated code.
How to contribute:
- Submit an issue describing the problem clearly
- Assign it to @copilot with your requirements
Add this marketplace to Claude Code:
/plugin marketplace add antvis/chart-visualization-skillsOr you can directly install the skills for your multiple agents:
npx skills add antvis/chart-visualization-skills- 📊 chart-visualization: A comprehensive chart generation skill powered by AntV that provides 26+ chart types for intelligent data visualization.
Chart Visualization intelligently selects the most appropriate chart type from 26+ available options, extracts parameters based on detailed specifications, and generates high-quality chart images. It covers time series, comparisons, part-to-whole, relationships, geographic, hierarchical, statistical, and specialized visualizations.
- 🎨 infographic-creator: Create beautiful infographics based on given text content. Use when users request to create infographics.
Infographic Creator uses AntV Infographic to transform data, information, and knowledge into a perceptible visual language. It combines visual design with data visualization, providing 50+ templates including lists, sequences, hierarchies, comparisons, relations, and charts. It compresses complex information with intuitive symbols to help audiences quickly understand and remember key points.
- 🖼️ icon-retrieval: Search and retrieve icon SVG strings from icon library. Returns up to 5 matching icons by default (customizable).
Icon Search helps users find appropriate icons for various use cases including infographics, web development, design, and more. Search by keywords to discover available icons and retrieve their SVG strings directly. Each search returns up to 5 matching icons by default (customizable via topK parameter) with their URLs and complete SVG content.
- 📝 narrative-text-visualization: Generate structured narrative text visualizations from data using T8 Syntax.
Narrative Text Visualization (T8) transforms unstructured data into semantically rich narrative reports using T8 Syntax - a declarative Markdown-like language for creating data narratives with entity annotations. It's LLM-friendly and framework-agnostic, working seamlessly with HTML, React, and Vue. Perfect for creating data analysis reports, summaries, and insights documents with entities like metrics, values, trends, and dimensions properly labeled. Features include built-in mini charts, standardized styling, and professional formatting. Supports authentic data sources and provides lightweight, technology-agnostic rendering.
- 📋 antv-s2-expert: S2 multi-dimensional cross-analysis table development assistant. Use when users need help with S2 pivot tables, table sheets, or any @antv/s2 related development.
AntV S2 Expert helps users develop with the S2 multi-dimensional cross-analysis table engine. It provides comprehensive guidance on @antv/s2 core engine, @antv/s2-react and @antv/s2-vue framework bindings, @antv/s2-react-components advanced analysis components, and @antv/s2-ssr server-side rendering. Covers pivot tables, table sheets, custom cells, theming, events, interactions, sorting, totals, tooltips, frozen rows/columns, icons, pagination, and more.
- 🕸️ antv-g6-graph: G6 v5 graph visualization code generator. Use when users need to generate G6 graphs — network graphs, tree graphs, flow charts, mind maps, and any relational or graph-structured data visualization with the G6 library.
AntV G6 Graph generates accurate, runnable G6 v5 code following best practices. It covers core graph initialization, data structures (nodes, edges, combos), 10+ layout algorithms (force, dagre, circular, grid, mindmap, fishbone, etc.), all built-in node/edge/combo types, state management, 15+ behaviors (drag-canvas, zoom-canvas, click-select, lasso, etc.), 10+ plugins (minimap, tooltip, toolbar, legend, timebar, etc.), custom element development, transforms, and animation. Built-in guard rails prevent common v4→v5 migration pitfalls such as using deprecated G6.Graph() constructors, graph.data() APIs, or Mode-based behavior configuration.
- 📈 antv-g2-chart: G2 v5 chart code generator. Use when users need to generate G2 charts — bar charts, line charts, pie charts, scatter plots, area charts, heatmaps, and any statistical data visualization with the G2 library.
AntV G2 Chart generates accurate, runnable G2 v5 code following Spec Mode best practices. It covers 30+ chart types (interval, line, area, point, rect, cell, treemap, sankey, chord, wordCloud, gauge, and more), data transforms (stackY, dodgeX, binX, fold, etc.), coordinate systems (cartesian, polar, theta, radial), scales, interactions (brush, slider, legend filter), components (axis, legend, tooltip, annotation), and multi-view compositions. Built-in guard rails prevent common v4→v5 migration pitfalls such as using deprecated chain APIs, invalid palette names, or referencing d3 in user code.
Evaluation Results
Harness Engineering approach has been rigorously tested against 174 chart generation test cases, demonstrating significant improvements over baseline methods:
| Model | G2 | G6 |
|---|---|---|
| qwen3-coder-480b-a35b-instruct | ✨ 98.2% +17.7% | 94.8% +15.6% |
| Kimi-K2.5 | 97.7% +17.2% | 96.9% +17.7% |
| GLM-5.1 | 93.6% +13.1% | 92.8% +13.6% |
| DeepSeek-V3.2 | 90.8% +10.3% | ✨ 97.9% +18.7% |
| Context7 Baseline | 80.5% | 79.2% |
The results show that Harness Engineering enables LLMs to achieve near-production-ready accuracy (up to 98.2%), significantly outperforming the Context7 baseline approach.
Tip
More skills are coming soon.
It can be used as a library in your Node.js projects with CLI and API.
We also provide a CLI tool named antv for easy usage in your terminal, Install globally:
npm install -g @antv/chart-visualization-skillsRetrieve skills by query:
# Retrieve skills by query (metadata only)
antv retrieve "bar chart" --library g2 --topk 10
# Retrieve skills with full markdown content
antv retrieve "bar chart" --library g2 --content
# Retrieve skills and output as JSON
antv retrieve "bar chart" --library g2 --output jsonUsage for the command:
Usage: antv [options] [command]
CLI tool for AntV chart visualization skills retrieval
Options:
-V, --version output the version number
-h, --help display help for command
Commands:
retrieve [options] <query> Search for skills matching a query
help [command] display help for command
Options for retrieve:
--library <lib> Filter by library (e.g. g2, g6, x6)
--topk <n> Number of results to return (default: 7)
--strategy <s> Retrieval strategy: hybrid | vector (default: hybrid)
--content Include markdown content body in results
--output <format> Output format: json | text (default: "text")
Note: Constraints docs (category
__constraints__) are indexed as regular skill documents and appear naturally in search results when the query matches their content.
import { retrieve } from '@antv/chart-visualization-skills';
// With full markdown content (defaults)
const skills = retrieve('bar chart', { library: 'g2', topK: 5 });
// Metadata only (no content body)
const skills = retrieve('bar chart', { library: 'g2', topK: 5, content: false });
// With token budget — content trimmed to fit 4000 tokens, summary + code only
const skills = retrieve('bar chart', {
library: 'g2',
maxTokens: 4000,
progressiveLevel: 1,
});retrieve(query: string, options?: RetrieveOptions): Promise<Doc[]>
interface RetrieveOptions {
library?: string; // Library filter, e.g. 'g2' or 'g6'
topK?: number; // Number of results (default: 7)
content?: boolean; // Include markdown content body (default: true)
strategy?: 'hybrid' | 'vector'; // Retrieval strategy (default: 'hybrid')
maxTokens?: number; // Token budget — content trimmed to fit when set
progressiveLevel?: 0 | 1 | 2; // 0=full, 1=summary+code, 2=summary (default: 1)
}| Option | Type | Default | Description |
|---|---|---|---|
library |
string |
all | Library filter (g2 / g6 / x6) |
topK |
number |
7 |
Number of results |
content |
boolean |
true |
Include markdown content body |
strategy |
'hybrid' | 'vector' |
'hybrid' |
Retrieval strategy |
maxTokens |
number |
— | Token budget; content trimmed to fit when set |
progressiveLevel |
0 | 1 | 2 |
1 |
0=full, 1=summary+code, 2=summary only |
Notes:
- Default retrieval uses hybrid strategy: zvec native FTS (jieba) + Vector (HNSW ANN) + RRF fusion.
strategy: 'vector'uses pure ANN vector similarity search.content: truereturns markdown content body (frontmatter metadata is excluded).- Constraints docs (category
__constraints__) are indexed as regular skill documents and appear naturally in search results.- When
maxTokensis set, skill content is formatted and trimmed to fit the budget according toprogressiveLevel.
MIT License - see the LICENSE file for details.