moc-curator

name	moc-curator
description	Curate wiki maps and clustering pages for the LLM Wiki. Use when asked to "curate MOCs", "update maps", "find clusters", "what MOCs need updating", or "organize my notes".
allowed-tools	Read, Write, Edit, Bash, Glob, Grep, AskUserQuestion

Wiki Map Curator

This skill uses semantic embeddings to suggest map-page updates, discover new map opportunities, and improve the public wiki's navigational layer.

Capabilities

1. Find MOC Gaps: Notes that should be in existing MOCs but aren't linked
2. Discover Clusters: Groups of orphan notes that could form new MOCs
3. Detect Hub Notes: Notes with many outgoing links that could become MOCs
4. Per-Note Suggestions: Which MOCs a specific note should join

Embeddings include title, summary, and first 500 characters of body content for improved semantic accuracy.

Workflow

Phase 1: Gather Context

Run the clustering script to analyze the knowledge base:

python3 .claude/skills/moc-curator/scripts/cluster-notes.py --mode=full

Parse the JSON output:

• moc_updates: Existing MOCs with missing notes
• new_clusters: Orphan note clusters that could be MOCs

Phase 2: Present MOC Update Suggestions

For each MOC with gaps, show:

### [[moc-slug]] - MOC Title

Currently has X members. Suggested additions:

1. **[[note-slug]]** (score: 0.82)
   - Type: article | Tags: tag1, tag2
   - Shares tags: shared-tag

2. **[[another-note]]** (score: 0.75)
   - Type: youtube | Tags: tag3
   - High semantic similarity to existing members

Phase 3: Present New MOC Opportunities

For each orphan cluster:

### Potential New MOC: "Inferred Theme"

Common tags: tag1, tag2, tag3

Notes in this cluster:

- [[note-a]] - Note A Title
- [[note-b]] - Note B Title
- [[note-c]] - Note C Title

**Suggested MOC name**: `theme-name-guide.md`

Phase 4: User Selection (Blocking Gate)

Use the AskUserQuestion tool to ask which suggestions to apply:

question: "Which MOC suggestions would you like me to apply?"
header: "MOC Updates"
multiSelect: false
options:
  - label: "Apply all MOC updates"
    description: "Add suggested notes to existing MOCs"
  - label: "Apply all new MOC creations"
    description: "Create new MOCs from discovered clusters"
  - label: "Select specific suggestions"
    description: "Choose individual suggestions to apply"
  - label: "Skip all"
    description: "Don't make any changes"

Wait for explicit approval before making any changes.

Phase 5: Apply Changes

For MOC Updates:

1. Read the existing MOC file
2. Append new links to a ## Suggested section at the end

## Suggested

- [[new-note-1]] - Brief description of why it belongs
- [[new-note-2]] - Brief description of why it belongs

For New MOCs:

1. Read the writing-style skill for prose guidelines:
   • Use the Read tool on .claude/skills/writing-style/SKILL.md
   • Apply its 10 rules when writing MOC descriptions
2. Create a new map note following the pattern:

---
title: "Theme Name Guide"
type: map
tags:
  - common-tag-1
  - common-tag-2
---

Description of what this map covers.

## Section Name

- [[note-a]] - What this note contributes
- [[note-b]] - What this note contributes

Phase 6: Quality Check

Run linter and type check to catch any issues:

vp run check

If errors are found, fix them before completing the task.

Script Usage

# Full analysis (MOC gaps + new clusters + hub notes)
python3 .claude/skills/moc-curator/scripts/cluster-notes.py --mode=full

# Only MOC gaps
python3 .claude/skills/moc-curator/scripts/cluster-notes.py --mode=moc-gaps

# Only orphan clusters
python3 .claude/skills/moc-curator/scripts/cluster-notes.py --mode=new-clusters

# Only hub notes (potential MOCs)
python3 .claude/skills/moc-curator/scripts/cluster-notes.py --mode=hub-notes

# Suggestions for a specific note (used by adding-notes hook)
python3 .claude/skills/moc-curator/scripts/cluster-notes.py --mode=for-note --note=slug-name

# Adjust threshold (default: 0.7 - strict)
python3 .claude/skills/moc-curator/scripts/cluster-notes.py --threshold=0.6

# Adjust minimum links for hub detection (default: 5)
python3 .claude/skills/moc-curator/scripts/cluster-notes.py --mode=hub-notes --min-links=3

JSON Output Format

{
  "moc_updates": [
    {
      "moc": "slug",
      "moc_title": "Title",
      "current_members": 5,
      "missing_notes": [
        {
          "slug": "note",
          "title": "Title",
          "score": 0.82,
          "shared_tags": ["tag"],
          "type": "article"
        }
      ]
    }
  ],
  "new_clusters": [
    {
      "cluster_id": 0,
      "theme": "Inferred Theme",
      "common_tags": ["tag1", "tag2"],
      "notes": [{ "slug": "note", "title": "Title", "type": "article" }],
      "size": 3
    }
  ],
  "orphan_stats": {
    "total_orphans": 20,
    "clustered": 8,
    "unclustered": 12
  },
  "potential_mocs": [
    {
      "slug": "note-slug",
      "title": "Note Title",
      "type": "article",
      "outgoing_links": 7,
      "links": ["link-a", "link-b", "link-c"]
    }
  ],
  "for_note": [
    {
      "moc": "slug",
      "moc_title": "Title",
      "score": 0.82,
      "reason": "shares tags: x; semantic similarity: 82%"
    }
  ]
}

Error Recovery

Issue	Resolution
No embeddings cache	Script will auto-generate embeddings
sklearn not installed	pip install scikit-learn
Empty results	Try lower threshold: --threshold=0.6
sentence-transformers missing	pip install sentence-transformers

Quality Guidelines

• Only suggest notes with similarity >= 0.7 (strict threshold)
• New MOCs require at least 3 notes in a cluster
• Always show the reasoning (shared tags, similarity score)
• Let the user decide what to apply - never auto-modify MOCs