Data Dashboard logo

Data Dashboard

CommunityPopular
plugin87
data-dashboard

Build a dense, data-heavy screen - an analytics console or a trading terminal - with real charts drawn from tokens: candlestick, volume, depth, stacked area, waterfall, scatter, correlation matrix, donut, gauge, sparklines, heatmaps and order books. Use when the request is a dashboard, terminal, monitoring view, data console, or any screen whose job is to show many numbers at once. Covers the layout sequence, the SVG geometry, the accessibility contract per chart type, and the contrast traps that only appear in dark mode.

Overview

Publisherplugin87
Repositoryux-ui-agent-skills
Skill namedata-dashboard
Stars
3K
Forks
290
Bundled files
Instructions only
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by plugin87 on GitHub. Read the source before you install it.

Installation

Install the Data Dashboard AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/plugin87/ux-ui-agent-skills/tree/main/.claude/skills/data-dashboard
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/plugin87/ux-ui-agent-skills.git /tmp/ux-ui-agent-skills
mkdir -p .claude/skills
cp -r /tmp/ux-ui-agent-skills/.claude/skills/data-dashboard .claude/skills/data-dashboard
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Data Dashboard in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Data Dashboard on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Data Dashboard is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Skill: Data dashboard

Step 0: is the kit here? This skill reads files from the kit. Check once: ls ${CLAUDE_SKILL_DIR}/../../../tokens >/dev/null 2>&1 && echo KIT_OK || echo KIT_MISSING On KIT_MISSING only the skill folders were installed, which is what npx skills add does. Say so plainly, point the user at npx ux-ui-agent-skills init or the plugin install, and stop. Do not guess the contents of a file you could not open.

A dashboard is the hardest screen to keep honest: it is dense, it is mostly non-text, and almost every default makes it worse. This is the recipe that produced ${CLAUDE_SKILL_DIR}/../../../examples/showcase/ and ${CLAUDE_SKILL_DIR}/../../../examples/terminal/, both of which pass every gate in both themes.

1. Decide the brief before drawing anything

Answer these in one line each, or /grill-me them out of the user:

  • Who reads this, and what decision do they make from it? A revenue console for an operator and a trading desk for a dealer look nothing alike.
  • What is the one number that leads? A dashboard without a lead is a wall.
  • What is the update cadence? Hourly figures get a "last run" line. Live figures get a tape and a timestamp per row.
  • How dense is honest? A terminal earns 12 panels. A weekly report does not.

2. Layout sequence, not a grid of equal cards

Four equal stat cards is the single loudest tell that nobody designed this. Compose in bands, each a different shape:

BandContentWhy
1Hero metric at 3x body size + its own chart, spanning most of the widthThe eye lands once
2Two or three secondary figures, stacked or narrowQuiet by construction
3Three different chart shapes side by side (donut, ranked bars, sparklines)Variety reads as intent
4A table that really sorts, plus a feed or logDetail after summary
5A footer line that ends the pageA page must end on purpose

Use a 12-column grid with explicit spans (span 8 / span 4), collapsing to 6 then 1. Never repeat(auto-fit, …) for the primary bands - it produces exactly the equal-card wall you are avoiding.

3. The chart inventory

All of these are plain inline SVG with viewBox, sized by CSS, filled from --color-chart-*. No chart library, no runtime dependency.

ChartGeometryUse it for
Bar / column<rect> per bucket, rx for a soft cornerDiscrete periods
Candlestick<line> wick + <rect> body per period, class by up/downOHLC price
VolumeBars under the price chart, same x-scale, opacity:.45Conviction behind a move
Moving average<polyline> over the candles, 2px, round joinsTrend through noise
Line / sparkline<polyline>, pathLength="1" for a draw animationTrend in a cell
Area / stacked area<path> per band: forward along the top, reverse along the previous baseline, closeComposition over time
DepthTwo cumulative area paths mirrored around the midBook liquidity
WaterfallFloating <rect> per step + dashed connectors between levelsAttribution
Donut<circle> with stroke-dasharray and rotate() per segmentShare of a whole, 3-5 slices
Radial gaugeSame trick, stroke-linecap:round, 270-degree sweepOne value against a limit
Scatter / bubble<circle>, radius = third dimensionTwo measures plus weight
Correlation matrixHTML table, cell tint via color-mixPairwise relationships
HeatmapHTML table or <rect> grid, tint by valueCohorts, calendars
Order bookHTML table with a depth bar per rowBids and asks

Generate the geometry, do not hand-place it. Write a small deterministic script (a seeded walk) that emits the SVG elements, then paste the output in. Hand-typed coordinates drift and cannot be regenerated.

4. Domain correctness is part of craft

Gates cannot catch a chart that is drawn beautifully and means nothing. Check by hand:

  • A correlation matrix is symmetric and its diagonal is exactly 1. Random values in both triangles is an instant tell to anyone who reads one for a living.
  • A stacked area sums to the total; bands must not cross or leave gaps. The usual bug is reversing the baseline by index instead of by coordinate, which shows as dark wedges between bands.
  • A waterfall needs connectors, and the closing bar must equal opening plus the steps.
  • Depth rises away from the mid on both sides; it never falls.
  • A donut whose segments do not total the whole is a pie chart lying.

5. Accessibility contract, per chart

  • Every SVG gets role="img" plus either aria-label or <title>+<desc> referenced by aria-labelledby. Say what the chart shows, not that it is a chart: "Range 182.06 to 200.10, closing at 187.84" beats "price chart".
  • Never encode meaning in colour alone. Green with a + and an arrow; red with a -. Up and down also carry a word in the table cell.
  • Tables get <caption> (use .sr-only when the panel header already says it), scope="col" / scope="row", and aria-sort only if clicking really sorts.
  • Sortable headers, range tabs and theme toggles all declare state - so they must change it on a real click. verify_interactive.mjs clicks them and checks.

6. The traps that only appear when you measure

Every one of these was found on a screen that looked finished. The first group a gate caught; the ones marked "no gate sees it" were caught by opening the screenshot and looking, which is why that step is not optional:

  • A chart palette is not a text palette. White initials on --color-chart-4 measured 2.65:1 in dark. Chart colours are tuned for large filled shapes.
  • Success green as text on a raised surface measured 4.43:1 - four hundredths short of AA. Put the colour on the icon and the sign; keep the text neutral.
  • axe must not run mid-transition. Disable transitions before flipping the theme, or contrast is sampled while colours are still moving.
  • Dense tables do not fit a phone. Drop the column a phone can lose - the sparkline, the derived value - in a media query, and say so in a comment. Do not shrink type below the scale.
  • Stagger comes from a token: calc(var(--duration-fast) * 0.6 * n), never a typed millisecond.
  • A grid row stretches its items. Two fields side by side where only one has a hint: the hint-less control grows to match the taller column and the two inputs end up different heights. Set align-content: start on the field and an explicit block-size on the control. Measure the heights in the render - the difference is obvious in a screenshot and invisible in the markup.
  • The same stretch hits whole panels, and no gate sees it. A row is as tall as its tallest panel, and the short one packs its content to the top and ends in a bordered void - a heatmap panel with 130px of nothing under the caption, a right column that stops two panels short of the map beside it. Every panel in a row has to say which it is: one child absorbs the slack (grid-template-rows:auto minmax(0,1fr), and that child's own tracks go minmax(<min>,1fr) so its rows grow too), or the rows share it out (align-content:space-between), or the short column earns another panel. Screenshot the row and look at the bottom edge of each panel; this never shows up in the markup.
  • auto-fit silently drops a column, and no gate sees it. Three gauges in a panel about 340px wide with minmax(min(100%,7rem),1fr) resolved to two tracks and orphaned the third beside an empty cell. When the count is fixed and meaningful, write repeat(3,minmax(0,1fr)) and give narrow widths their own rule.
  • A category chart needs its categories on the page, and no gate sees it. A Pareto of five defect types whose names live only in the aria-label is a row of anonymous bars to everyone looking at it. Put the labels under the bars - lay the plot out on an even pitch inside its viewBox so a matching grid of labels lines up without a second coordinate system.
  • An .sr-only needs a positioned ancestor. Inside a scroller it otherwise resolves against the initial containing block, lands outside the viewport, and inflates the document's scroll width by hundreds of pixels.
  • A zero-area element that only has area while it animates reads as content lost under prefers-reduced-motion. Give a radar sweep a wedge, not a 1px line.
  • Never reuse a class name across layout and colour. .c4 as both a four-column span and the fourth chart colour painted an entire panel pink.

7. The loop

bash
node ${CLAUDE_SKILL_DIR}/../../../scripts/gate.mjs measure_render.mjs <file> && node ${CLAUDE_SKILL_DIR}/../../../scripts/gate.mjs measure_render.mjs --dark <file>
node ${CLAUDE_SKILL_DIR}/../../../scripts/gate.mjs verify_states.mjs <file>  && node ${CLAUDE_SKILL_DIR}/../../../scripts/gate.mjs axe_audit.mjs --dark <file>
node ${CLAUDE_SKILL_DIR}/../../../scripts/gate.mjs verify_responsive.mjs <file> --scale=1.25
node ${CLAUDE_SKILL_DIR}/../../../scripts/gate.mjs verify_interactive.mjs <file>
DS_REQUIRE_BROWSER=1 node ${CLAUDE_SKILL_DIR}/../../../scripts/slop_tells.mjs --strict <file> && DS_REQUIRE_BROWSER=1 node ${CLAUDE_SKILL_DIR}/../../../scripts/taste_audit.mjs --strict <file>

Then screenshot it and look, in both themes, at 1600 and 390 wide. The gates will pass a chart that is drawn wrong. A correlation matrix that is not symmetric is 16/16 green and still embarrassing.

Worked references, both gate-verified: ${CLAUDE_SKILL_DIR}/../../../examples/showcase/index.html (revenue console) and ${CLAUDE_SKILL_DIR}/../../../examples/terminal/index.html (trading desk, ten chart types).

Frequently asked questions

What does the Data Dashboard AI skill do?

Build a dense, data-heavy screen - an analytics console or a trading terminal - with real charts drawn from tokens: candlestick, volume, depth, stacked area, waterfall, scatter, correlation matrix, donut, gauge, sparklines, heatmaps and order books. Use when the request is a dashboard, terminal, monitoring view, data console, or any screen whose job is to show many numbers at once. Covers the layout sequence, the SVG geometry, the accessibility contract per chart type, and the contrast traps that only appear in dark mode.

Why use Data Dashboard on TypingMind?

Because you install it once and use it with any model. Data Dashboard is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Data Dashboard in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/plugin87/ux-ui-agent-skills/tree/main/.claude/skills/data-dashboard. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Data Dashboard?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Data Dashboard?

As many as you like. As long as a model supports skills, you can use Data Dashboard with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Data Dashboard AI skill free?

Yes. It is published on GitHub by plugin87 under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇