update_vault_index

📇 Update Vault Index Skill

Automatically sync .claude/vault_index.md with your actual MOCs/, Projects/, and Reading/ directories.

PURPOSE

The vault index is the entry point to your research graph (per CLAUDE.md rules). It must stay current to reflect:

• All MOCs (research areas)
• All projects categorized by status (Seed/Idea/Active/Writing/Finished)
• Quick access to recent papers

WORKFLOW

0. Read last-run timestamp

Read .claude/vault_maintanance.md and note the update_vault_index: timestamp.

Default: incremental scan — only process files modified after the update_vault_index: timestamp. Use find Projects/ MOCs/ -newer <reference_file> to identify changed files. To use a reference file: touch -t YYYYMMDDHHMI /tmp/vault_index_ref && find ... -newer /tmp/vault_index_ref

• If no files were modified: skip the scan and report "Index already up to date."
• Only run a full rescan if the user explicitly requests it or no timestamp exists yet.
• Always do a full scan of Projects/ for status categorization even in incremental mode (status may have changed without file modification time changing).

1. Scan vault structure

Scan directories:

• MOCs/ → extract all MOC file names
• Projects/ → extract all project file names + status field
• Reading/ → optional: list recent papers (last 5-10)

Extract metadata:

• For each project: read YAML frontmatter to get status field
• For each MOC: extract 1-line description if present

2. Categorize projects

Group projects by status (maintain this order):

• Seed: status = seed (early-stage ideas generated by Claude)
• Idea: status = idea (refined projects approved by human)
• Active: status = active (actively working on it)
• Writing: status = writing (writing paper and finishing project)
• Finished: status = finished (completed)

Sort alphabetically within each group.

3. Build index sections

Structure (preserve exactly):

# Vault Index

## Research Areas (MOCs)
- [[MOC Name]] - brief description (1 line)
- ...

## Seed Projects
- [[Project Name]] - brief description (1 line)
- ...

## Idea Projects
- [[Project Name]] - brief description (1 line)
- ...

## Active Projects
- [[Project Name]] - brief description (1 line)
- ...

## Writing Projects
- [[Project Name]] - brief description (1 line)
- ...

## Finished Projects
- [[Project Name]] - brief description (1 line)
- ...

Get descriptions from:

• MOC files: first non-heading paragraph or subtitle
• Project files: "Seed Idea" or "Current Working Idea" field
• Keep descriptions to 1 line max

4. Generate minimal descriptions

Rules:

• Extract from actual file content (don't invent)
• Max 1 line (70 chars)
• Use present tense
• Keep factual, avoid marketing language

Example:

- [[Persona Fingerprinting to Prevent Emergent Misalignment]] -
  Preserve identity fingerprint during finetuning to prevent misalignment

5. Update vault_index.md

• Replace entire file (sections only, no other changes)
• Preserve markdown formatting
• Verify all wiki links use exact file names
• Verify no broken links

6. Update vault_maintanance.md

After writing vault_index.md:

• Run date via Bash to get the current system time
• Read .claude/vault_maintanance.md
• Update update_vault_index: line to current date and time in format DD-MM-YYYY HH:MM
• Reset papers_since_vault_index: to 0
• Reset projects_since_vault_index: to 0
• Write the updated file

Example update:

update_vault_index: 10-02-2026 14:35

7. Verify output

✅ All MOCs present with descriptions ✅ All seed projects listed (status: seed) ✅ All idea projects listed (status: idea) ✅ All active projects listed (status: active) ✅ All writing projects listed (status: writing) ✅ All finished projects listed (status: finished) ✅ Projects in correct order (Seed → Idea → Active → Writing → Finished) ✅ Alphabetical within each section ✅ Wiki links match exact file names in [[brackets]] ✅ No orphan files referenced

TOKEN EFFICIENCY

Optimizations:

• Read only YAML frontmatter + 1-2 lines per file
• Use grep to extract status field: status:\s*
• Single pass through each directory
• No deep content analysis
• Use file listing instead of reading metadata separately

Fast execution:

• ~50-100 tokens for full vault with 20+ projects

OUTPUT RULES

• Minimal, structural updates only
• One-line descriptions max
• Bullet format only
• No prose explanation needed in output
• Example: "Index updated: 4 MOCs, 5 seed, 0 idea, 1 active, 0 writing, 0 finished"

SELF-CHECK

✅ Did I read all MOCs, projects from actual directories? ✅ Did I extract status field from each project file? ✅ Did I categorize correctly (Seed/Idea/Active/Writing/Finished)? ✅ Did I maintain correct section order (Seed → Idea → Active → Writing → Finished)? ✅ Did I use exact file names in [[wiki links]]? ✅ Did I get descriptions from actual file content? ✅ Did I keep descriptions to 1 line max? ✅ Did I verify no broken links in output? ✅ Did I preserve vault_index.md structure exactly? ✅ Did I update vault_maintanance.md with current timestamp?