AgentSkillsCN

graph-builder

在Obsidian vault中分析并构建知识图谱链接。查找孤立笔记,推荐关联节点,添加反向链接,可视化链接结构。可通过“/graph”、“分析链接”、“查找孤立节点”、“推荐关联节点”等短语触发。

SKILL.md
--- frontmatter
name: graph-builder
description: Analyze and build knowledge graph links in Obsidian vault. Find orphan notes, suggest connections, add backlinks, visualize link structure. Triggers on /graph, "analyze links", "find orphans", "suggest connections".

Graph Builder

Analyze vault link structure and build meaningful connections between notes.

Use Cases

  1. Analyze — Statistics and insights about vault graph
  2. Find Orphans — Notes without incoming/outgoing links
  3. Suggest Links — AI-powered connection recommendations
  4. Add Links — Batch link creation based on content analysis
  5. Visualize — Export graph data for visualization

Quick Commands

CommandAction
/graph analyzeFull vault analysis with stats
/graph orphansList unconnected notes
/graph suggestGet link suggestions
/graph addApply suggested links

Analysis Output

code
📊 Vault Graph Analysis

Total notes: 247
Total links: 892
Orphan notes: 12
Most connected: [[MEMORY]] (47 links)
Weakest domain: learnings/ (avg 1.2 links/note)

🔗 Suggested connections:
• [[Project A]] ↔ [[Client X]] (mentioned 5x)
• [[Idea B]] → [[MOC/Ideas]] (category match)

Domain Configuration

Domains are configured in references/domains.md. Default structure:

  • daily/ — Daily journal entries
  • thoughts/ — Processed ideas, reflections, learnings
  • goals/ — Goal cascade files
  • MOC/ — Maps of Content (index pages)
  • projects/ — Project notes

Link Building Strategy

  1. Entity extraction — Find mentions of existing notes
  2. Category mapping — Connect notes to relevant MOCs
  3. Temporal links — Link daily entries to related thoughts
  4. Cross-domain — Bridge domains (project ↔ goal ↔ daily)

Scripts

  • scripts/analyze.py — Graph statistics and orphan detection
  • scripts/add_links.py — Batch link insertion

References

  • references/domains.md — Domain definitions and rules
  • references/frontmatter.md — Frontmatter schema for notes

Output Format

Reports use plain markdown (for vault notes) or HTML (for Telegram).

For vault: Standard markdown with [[wiki-links]] For Telegram: HTML tags (b, i, code only)