obsidian:bulk-scan

Bulk Scan

Efficiently collect metadata for many vault notes at once. Use this instead of calling obsidian wordcount/properties/tags per-file in a loop.

Strategies

1. File Listing (one call)

Get all files with metadata in a single call:

obsidian files ext=md format=json

Returns file paths. Use this as the base list for all subsequent operations.

2. Batch Content via qmd multi_get

Read multiple files at once instead of N individual obsidian read calls:

mcp__qmd__multi_get with:
- pattern: "folder/*.md"     # glob pattern
- maxLines: 20               # first 20 lines (frontmatter + intro)
- maxBytes: 10240

Or comma-separated list:

mcp__qmd__multi_get with:
- pattern: "file1.md,file2.md,file3.md"

3. Batch Properties via obsidian CLI

Get vault-wide property statistics in one call:

obsidian properties counts sort=count format=json   # all properties
obsidian tags counts sort=count format=json          # all tags

Per-folder:

obsidian files folder=<path> ext=md   # list files in folder

Then batch-read frontmatter via qmd:

mcp__qmd__multi_get with:
- pattern: "<folder>/*.md"
- maxLines: 10   # frontmatter only

4. Link Statistics

Bulk link health is available in single calls:

obsidian orphans          # all orphans at once
obsidian deadends         # all deadends at once
obsidian unresolved       # all unresolved at once

No need to check per-file.

5. Batch Semantic Search

When checking many notes against the vault (e.g., consistency-checker), batch qmd queries:

# Search for each note's topic — but do it sequentially via mcp__qmd__query
# The MCP server keeps the model in memory, so repeated queries are fast

qmd MCP maintains the embedding model in memory. Sequential queries are fast (no cold start per query).

When to Use What

Need	Bad Pattern (slow)	Good Pattern (fast)
All file paths	Loop obsidian file per name	obsidian files ext=md format=json (1 call)
Read N files	Loop obsidian read per file	mcp__qmd__multi_get with glob/list
Frontmatter of N files	Loop obsidian properties file=X	mcp__qmd__multi_get maxLines=10 (frontmatter only)
Orphans/deadends	Loop obsidian links per file	obsidian orphans / obsidian deadends (1 call)
Word count of N files	Loop obsidian wordcount per file	Read files via mcp__qmd__multi_get, count words in content
Search N topics	N separate web searches	Sequential mcp__qmd__query (MCP model stays warm)

Combining with vault-git

For incremental operations:

1. Get changed files via vault-git skill (git diff)
2. Batch-read those files via mcp__qmd__multi_get with comma-separated list
3. Process in memory

This avoids scanning unchanged files entirely.