What is the building tables skill?
Building tables is a skill in ancoleman/ai-design-components. It guides an AI coding agent to build tables and data grids by data volume: semantic HTML under 100 rows, client features up to 1,000, server operations beyond that, virtual scrolling past 10,000, with accessibility and responsive patterns included.
Building tables is one skill inside ai-design-components, a collection the README describes as 76 skills for frontend, backend and DevOps work. This one covers tables and data grids, from a simple read-only list to a grid with millions of rows.
Its central idea is a decision framework based on data volume. The agent first asks how many rows and which features you need, then picks the lightest tier that works instead of reaching for a heavy grid by default.
The five tiers by row count
- Under 100 rows: a semantic HTML table with progressive enhancement.
- 100 to 1,000 rows: client-side sorting, filtering and pagination.
- 1,000 to 10,000 rows: server-side operations with API pagination.
- 10,000 to 100,000 rows: virtual scrolling with windowing.
- Over 100,000 rows: an enterprise grid with streaming and workers.
Who the building tables skill is for
Best for
- Admin screens and dashboards that list records with sort, filter and bulk actions.
- Teams choosing between a headless table library and a full grid.
- Tables that must pass keyboard and screen reader checks.
- Wide tables that need a plan for phones.
Not for
- Charts and visualizations, which belong to a different skill.
- Layout tables for page design; this is about tabular data.
- A fixed visual style. It sets structure and behaviour, and takes colours and spacing from your design tokens.
Use it with UiChemy
Tables built with this skill live in your app code. For marketing pages and content sites, UiChemy Composer writes HTML, CSS and JS for WordPress. Click Get Started on uichemy.com.
How to install Building Tables Skill: Data Grids for AI Coding Agents
git clone https://github.com/ancoleman/ai-design-components.git && cd ai-design-components && ./install.shThe README lists Claude Code CLI as a prerequisite and recommends the Context7 MCP for current library docs.Clone the repository
Run the clone command and move into the folder. The README names the interactive installer as its recommended route.
Run the installer
Run
./install.shand choose what to install from the menu../install.sh plugins listshows the available plugin groups.Or copy only this skill
The skill lives in
skills/building-tables. Copy that folder into your agent's skills directory if you want nothing else.Ask for a table
Describe the data, for example 20,000 orders with sort, filter and CSV export, and let the agent pick the tier.
Check that it loaded
Ask your agent to build a table for 50,000 rows. It should name a tier, mention virtual scrolling and suggest TanStack Table or AG Grid.
Key features
- Volume-based decision frameworkThe skill maps row count to an approach, so a 50-row list does not ship with a virtualized grid and a 500,000-row feed does not load into a plain table.
- Three implementation tiersBasic tables, interactive tables up to about 10,000 rows, and advanced grids beyond that, each with its own reference file and a matching example.
- Performance thresholdsClient operations stay under 1,000 rows at roughly 50 ms, server operations target under 200 ms API time, and virtual scrolling aims for 60 fps with constant memory.
- Sorting and filteringSingle or multi-column sorting, custom numeric, date and natural sorts, column filters, a global search and AND or OR filter logic, with keyboard support.
- Pagination choicesClient-side pagination for small sets, server-side for large ones, and infinite scroll as an alternative.
- Selection and bulk actionsSingle and multi-row selection, Shift-click range selection and a toolbar for bulk operations.
- Inline editing and exportCell or row editing with validation and optimistic updates, plus CSV, Excel and PDF export that streams large files.
- Accessibility rulesSemantic HTML, the ARIA grid pattern for interactive tables, full keyboard navigation and screen reader announcements, with a script to validate them.
- Four responsive strategiesHorizontal scroll, card stack, priority columns, and truncate with expand, each shown in a responsive patterns example.
- Library guidanceTanStack Table as the headless default for custom designs and AG Grid for feature-complete enterprise grids, with a comparison reference.
Frequently asked questions
What is the building tables skill?
It is a skill in ancoleman/ai-design-components that teaches an AI coding agent how to build tables and data grids. It selects an approach by row count, then covers sorting, filtering, pagination, selection, editing, export, accessibility and responsive behaviour.
How do I install the building tables skill?
Clone ancoleman/ai-design-components and run the interactive installer, ./install.sh, which the README recommends. You can also copy the skills/building-tables folder into your agent's skills directory. Then ask for a table and give the row count and the features you need.
Which table library does the skill recommend?
TanStack Table is the primary pick for custom designs: it is headless, TypeScript-first and about 15KB. AG Grid is the enterprise option for very large datasets and spreadsheet-like use, with a free community edition and a paid enterprise one.
When should I use virtual scrolling for a table?
The skill suggests virtual scrolling from about 10,000 rows, where windowing keeps rendering at 60 fps with constant memory. Below that, client-side or server-side pagination is simpler. Past 100,000 rows it moves to streaming and workers.
Does the skill cover table accessibility?
Yes. It calls for semantic table markup, the ARIA grid pattern for interactive tables, keyboard navigation and screen reader announcements. The skill folder includes a script, validate_accessibility.js, to check an implementation.
How does the skill handle tables on mobile?
It offers four strategies: horizontal scroll, which keeps the structure, a card stack that turns rows into cards, priority columns that hide less important ones, and truncate with expand for compact rows with detail on demand.
Is the building tables skill the same as a form skill?
No. It handles tabular data only. For input fields, errors and flow, see the form design skill. For checking a finished table with automated audits, A11y Ally runs WCAG checks.
What are building tables skill alternatives?
You can follow a library's own docs, or use a component spec skill such as Component Spec to write a table's API before building it. This skill adds the row-count decision framework and performance thresholds on top.


