Claude Code Plugins

Community-maintained marketplace

Feedback

Analyze the knowledge graph for insights. Use when asked to "analyze connections", "graph report", "show hubs", "find orphans", or "knowledge map".

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name exploring-graph
description Analyze the knowledge graph for insights. Use when asked to "analyze connections", "graph report", "show hubs", "find orphans", or "knowledge map".
allowed-tools Read, Glob, Grep

Exploring the Knowledge Graph

This skill analyzes the structure of connections in the knowledge base to surface insights.

Workflow

1. Build the Connection Map

Parse all wiki-links from every content file:

# Extract all wiki-links
grep -oh '\[\[[^]]*\]\]' content/*.md

# Get links per file
for f in content/*.md; do
  echo "=== $f ==="
  grep -o '\[\[[^]]*\]\]' "$f"
done

Build an adjacency list:

  • note-a → links to: [note-b, note-c]
  • note-b → links to: [note-a]
  • etc.

2. Calculate Metrics

For each note, determine:

Metric Description
Outgoing links Wiki-links in this note
Incoming links Other notes linking to this one
Total connections Outgoing + Incoming

3. Identify Patterns

Hub Notes (Most Connected)

Notes with the highest total connections - these are central to the knowledge base.

Orphan Notes

  • Full orphans: 0 incoming AND 0 outgoing links
  • Dead ends: Has outgoing but 0 incoming
  • Sources: Has incoming but 0 outgoing

Clusters

Groups of notes that link heavily to each other but less to the rest. Identify by:

  • Shared tags
  • Mutual links
  • Map membership - notes linked from a type: map note form visual clusters

Map Notes (MOCs)

Map notes (type: map) act as cluster centers on the graph:

  • Appear as pink hexagons instead of circles
  • Pull member notes toward them with gravitational force
  • Members are defined by wiki-links FROM the map
  • Use the Maps filter in the graph UI to focus on specific clusters

To find existing maps:

grep -l "type: map" content/*.md

Broken Links

Wiki-links pointing to non-existent notes.

4. Generate Insights Report

## Knowledge Graph Analysis

### Overview
- Total notes: 15
- Total connections: 42
- Average connections per note: 2.8

### Hub Notes (Top 5 Most Connected)
| Note | Outgoing | Incoming | Total |
|------|----------|----------|-------|
| [[central-concept]] | 5 | 8 | 13 |
| [[key-framework]] | 4 | 6 | 10 |
| ... | | | |

### Orphan Notes (Need Attention)
**Full Orphans** (isolated):
- `lonely-note.md` - no connections at all

**Dead Ends** (no incoming links):
- `new-note.md` - links out but not referenced

**Sources** (no outgoing links):
- `reference-only.md` - referenced but doesn't link

### Potential Clusters
**Vue Ecosystem** (5 notes, 12 internal links):
- [[vue-composables]]
- [[vue-testing]]
- [[nuxt-patterns]]
- ...

**AI/Agents** (3 notes, 6 internal links):
- [[12-factor-agents]]
- [[context-efficient-backpressure]]
- ...

### Connection Opportunities
Based on shared tags and titles, these notes might benefit from links:
- [[note-a]] and [[note-b]] share 3 tags but aren't linked

Graph Health Indicators

Indicator Healthy Warning
Orphan rate <10% >20%
Avg connections >2 <1
Broken links 0 Any

Quality Checklist

When analyzing:

  • Parsed all wiki-links from all files
  • Counted incoming/outgoing per note
  • Identified hub notes
  • Found orphan notes
  • Detected potential clusters
  • Checked for broken links
  • Suggested connection opportunities