agentsclimarketplace

Visualization datascape

Skill dmitrykrivaltsevich/claude-skills/skills/visualization-datascape

Generates immersive 3D cyberspace point-cloud visualizations from structured data. Renders an interactive cyberpunk cityscape with explorable data vaults, WASD movement, and orbit controls. Use when the user asks to visualize data as an immersive 3D scene, wants a cyberspace or cyberpunk datascape, asks for an interactive 3D dashboard, wants to explore data in a Blackwall-style environment, or wants to visualize an Obsidian vault as a 3D knowledge graph.From its SKILL.md

Install
npx -y skills add dmitrykrivaltsevich/claude-skills --skill visualization-datascape

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

8.5 KB, as published. Nobody here has run it

Visualization Datascape Skill

Contents

  1. Architecture
  2. Workflow
  3. Obsidian Vault Workflow
  4. JSON Config Format
  5. Running the Script
  6. LLM Responsibilities
  7. Visual Design
  8. Known Bugs to Avoid

Architecture

Script = data pipe. generate.py reads a JSON config from stdin, produces a self-contained HTML file to stdout. It handles templating, vault positioning, color assignment, and JS code generation.

LLM = brain. The LLM analyzes the user's data, decides how to partition it into vaults (up to 32,768; 3–8 recommended for readability), writes HTML content for each vault's info panel, picks stats/glyphs, and structures the JSON config.

Workflow

1. User provides data (text, table, report, URL content, etc.)
2. LLM analyzes → decides vault structure (3–8 recommended)
3. LLM builds JSON config with title, stats, vaults, glyphs
4. LLM writes JSON to a temp file
5. Run: uv run --no-config skills/visualization-datascape/scripts/generate.py -i /tmp/config.json -o ~/Downloads/Datascape.html
6. Open the HTML file in a browser

Obsidian Vault Workflow

When the user asks to visualize an Obsidian vault, use obsidian_to_datascape.py instead of manually building JSON:

# Step 1: Parse vault → JSON config
uv run --no-config skills/visualization-datascape/scripts/obsidian_to_datascape.py \
  "/path/to/obsidian/vault" \
  -o /tmp/obsidian-config.json

# Step 2: Generate HTML from config
uv run --no-config skills/visualization-datascape/scripts/generate.py \
  -i /tmp/obsidian-config.json \
  -o ~/Downloads/Obsidian-Knowledge-Graph.html

The parser automatically:

  • Creates one vault per .md note (excludes .obsidian/ directory)
  • Extracts [[wikilinks]] as connections between vaults
  • Embeds images from ![[image.png]] and ![](path) as base64 data URIs (up to 512KB each)
  • Shows PDF references, external URLs, tags, headings, and content previews
  • Colors vaults by folder
  • Temporal inference: notes with YYYY-MM-DD in the filename automatically get:
    • Synthetic month vaults (e.g. "JAN 2025") and year vaults (e.g. "2025")
    • Connections: dated note → month → year
    • Chronological prev/next connections between adjacent dated notes, months, and years

JSON Config Format

{
  "title": "MAIN TITLE",
  "subtitle": "optional subtitle line",
  "stats": [
    {"label": "data vaults", "value": "5"},
    {"label": "total items", "value": "1,200"}
  ],
  "vaults": [
    {
      "id": "unique-id",
      "name": "DISPLAY NAME",
      "html": "<div class=\"pt\">Title</div><div class=\"pd\">Content</div>",
      "color": "0x00ff66"
    }
  ],
  "glyphs": ["TERM_1", "code::ref", "data_key"],
  "connections": [
    {"from": "vault-id-1", "to": "vault-id-2"}
  ]
}

Required Fields

FieldTypeConstraint
titlestringNon-empty
vaultsarray1–32,768 entries (3–8 recommended)
vaults[].idstringUnique, non-empty
vaults[].namestringNon-empty, shown as label
vaults[].htmlstringHTML for the side panel

Optional Fields

FieldDefault
subtitleempty
statsnone shown
vaults[].colorauto-assigned from green palette
glyphsauto-generated from vault names
connectionsauto nearest-neighbor; array of {from, to} vault IDs for animated particle exchange

Panel HTML CSS Classes

Use these classes in vault html for consistent styling:

ClassPurpose
ptPanel title (large, green, uppercase)
psPanel subtitle (small, dim)
phSection heading (bordered)
pdData/paragraph line
pvHighlighted value (bold green)
pwSource/footnote (tiny, dim)
table, th, tdData tables
td.hHighlighted table cell

Media Classes

Images and videos in vault html get automatic click-to-lightbox behavior:

Class / PatternPurpose
img.piInline image (full panel width, click opens lightbox)
video.piInline video (plays in panel, double-click opens lightbox)
div.pi-deckHorizontal thumbnail strip (scrollable, each img inside is clickable)
div.pv-wrapSide-by-side layout: image + text (image left, text right)
data-full="url"Optional attribute on any img/video — lightbox shows this URL instead of the thumbnail src

Example vault HTML with media:

<div class="pt">Person Profile</div>
<div class="pv-wrap">
  <img class="pi" src="https://example.com/photo-thumb.jpg" data-full="https://example.com/photo-hires.jpg">
  <div><div class="pd">Name: <span class="pv">Jane Doe</span></div><div class="pd">Role: Engineer</div></div>
</div>
<div class="ph">Gallery</div>
<div class="pi-deck">
  <img src="https://example.com/img1-thumb.jpg" data-full="https://example.com/img1.jpg">
  <img src="https://example.com/img2-thumb.jpg" data-full="https://example.com/img2.jpg">
  <img src="https://example.com/img3-thumb.jpg">
</div>

Running the Script

From workspace root:

# From file
uv run --no-config skills/visualization-datascape/scripts/generate.py \
  -i /tmp/datascape_config.json \
  -o ~/Downloads/My_Datascape.html

# From stdin
cat config.json | uv run --no-config skills/visualization-datascape/scripts/generate.py \
  -o ~/Downloads/My_Datascape.html

Script has zero external dependencies — pure Python stdlib.

LLM Responsibilities

Vault Design (3–8 recommended)

  • Partition the user's data into logical clusters
  • Each vault = one coherent topic/category/dimension
  • Write rich HTML panels using the CSS classes above
  • Include tables for structured data, pv spans for key metrics

Stats Selection

  • Pick 3–5 top-level metrics for the HUD (top-right corner)
  • These should be the most impactful numbers

Glyph Selection

  • Pick 12–24 short code-like strings that float as ambient text
  • Mix domain terms, metrics, and cyberpunk flair
  • Keep each under 14 characters

Presentation Checklist

  1. Analyze user's data → identify 3–8 themes
  2. For each theme → write a vault with id, name, detailed HTML panel
  3. Pick 3–5 headline stats
  4. Pick 12–24 floating glyphs
  5. Set title + subtitle
  6. Generate JSON → run script → deliver HTML file

Visual Design

The output is a cyberpunk point-cloud cityscape with 3D hexagonal crystal lattice vault placement:

  • Crystal lattice: Vaults placed at vertices of a hexagonal close-packed (HCP) structure across 3 layers
    • Layer 0 (y=5): ground hex ring + center
    • Layer 1 (y=28): hex ring rotated 30° (HCP stagger — nestles in layer 0 hollows)
    • Layer 2 (y=50): apex points
  • Polyhedra by count: n=3 triangle, n=4 tetrahedron, n=5 bipyramid, n=6 hexagon, n=8 hex bipyramid, n=9+ two-layer crystal
  • Crystal edges: Vault connections trace nearest-neighbor bonds (like atomic bonds in a real crystal)
  • City: Dense point-cloud buildings in inner core (±80), sparse outer sprawl (±250)
  • Ground grid: Dot matrix extending ±350
  • Data streams: 250 falling green code columns
  • Particles: 25,000 ambient floating dust motes
  • Vaults: Pulsing sphere cores + expanding cube shells + vertical beams + labels
  • Glyphs: Floating code text sprites
  • Controls: Mouse orbit + WASD/arrows + space/shift for vertical + Q/E strafe

Known Bugs to Avoid

These apply to the Three.js template and must NEVER be reintroduced:

  • NEVER use EffectComposer or UnrealBloomPass — causes black screen silently
  • NEVER use Object.assign on position — Object3D.position is readonly. Use position.set() or position.copy()
  • Keep point sizes small: 0.25–0.8 range. Larger = washed out
  • Keep opacities low: 0.12–0.35 range. Higher = too bright
  • Tone mapping exposure: 0.7 is calibrated. Don't increase

Keep looking

Skills are one crate of 608,475. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.