Knowledge Graph Skill

Builds an interactive 3D knowledge graph from all markdown files in the workspace and displays it via Canvas.

How to Use

When the user asks to see a knowledge graph, visualize connections, or explore file relationships:

Step 1: Build the graph data

bash

python3 ~/clawd/skills/knowledge-graph/scripts/build-graph.py ~/clawd/ --output /tmp/graph-data.json

This scans all .md files recursively and extracts:

•Wiki-links ([[...]] references)
•Shared keywords (Jaccard similarity)
•Shared mentions (people, projects, technologies)

Step 2: Build the combined HTML with embedded data

bash

python3 ~/clawd/skills/knowledge-graph/scripts/build-graph.py ~/clawd/ \
  --output /tmp/graph-data.json \
  --embed ~/clawd/skills/knowledge-graph/assets/graph-viewer.html /tmp/knowledge-graph.html

The --embed flag takes the HTML template and injects the JSON data into it, producing a self-contained HTML file.

Step 3: Present via Canvas

code

canvas action=navigate url=file:///tmp/knowledge-graph.html

All-in-One Command

bash

python3 ~/clawd/skills/knowledge-graph/scripts/build-graph.py ~/clawd/ \
  --output /tmp/graph-data.json \
  --embed ~/clawd/skills/knowledge-graph/assets/graph-viewer.html /tmp/knowledge-graph.html

Then present:

code

canvas action=navigate url=file:///tmp/knowledge-graph.html

Alternative: Scan a Subdirectory

To graph only a specific project:

bash

python3 ~/clawd/skills/knowledge-graph/scripts/build-graph.py ~/clawd/projects/rhinomcp-dev/ \
  --output /tmp/graph-data.json \
  --embed ~/clawd/skills/knowledge-graph/assets/graph-viewer.html /tmp/knowledge-graph.html

Graph Features

•3D force-directed layout — nodes cluster by connections
•Color-coded by type: daily-note (blue), memory (green), task (orange), project (purple), knowledge (blue), config (grey), skill (red)
•Node size reflects file size + connection count
•Edge types: wiki-link (bright, particles), keyword (green, subtle), mention (orange, subtle)
•Search (top bar, or press /): filters by filename, keywords, mentions
•Type filters: click chips to show/hide types
•Click a node: opens detail panel with keywords, mentions, connected files
•Click connections in detail panel to navigate to that node
•Escape: close panel, clear search

Output Schema

The JSON output follows this schema:

json

{
  "nodes": [{ "id", "label", "type", "size", "keywords", "mentions", "lastModified" }],
  "edges": [{ "source", "target", "type", "weight" }],
  "stats": { "totalNodes", "totalEdges", "scannedAt" }
}

Requirements

•Python 3.10+ (stdlib only, no pip dependencies)
•Browser with WebGL for 3D visualization
•Internet connection for CDN (3d-force-graph library)