Agent skill
exploring-graph
Analyze the knowledge graph for insights. Use when asked to "analyze connections", "graph report", "show hubs", "find orphans", or "knowledge map".
Stars
163
Forks
31
Install this agent skill to your Project
npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/development/exploring-graph
SKILL.md
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:
bash
# 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: mapnote 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:
bash
grep -l "type: map" content/*.md
Broken Links
Wiki-links pointing to non-existent notes.
4. Generate Insights Report
markdown
## 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
Didn't find tool you were looking for?