cross-vault-wire

name	cross-vault-wire
description	Connect multiple vaults together via CLAUDE.md wiki path references. Use when the user wants one AI agent to access knowledge from another vault, or says 'wire my vaults together' or 'connect my executive assistant to my wiki'.
allowed-tools	Read Write Edit Glob Grep

Cross-Vault Wire

Connect multiple Obsidian vaults so AI agents in one project can access knowledge from another.

The Pattern

Each vault is independent, but any AI agent can reference any vault's wiki via CLAUDE.md path declarations. This is the "one source of truth, many consumers" pattern.

Workflow

Step 1: Identify the vaults

Ask the user:

• Which vault contains the knowledge? (the source vault)
• Which project needs access to that knowledge? (the consumer project)
• What kind of access is needed? (read-only query? full context? just recent?)

Step 2: Get the source vault path

The source vault needs:

• A wiki/index.md (or memory/wiki/index.md) — the master catalog
• Optionally a wiki/hot.md — recent context cache
• A clear folder structure that the consumer agent can navigate

Step 3: Add wiki path to consumer's CLAUDE.md

Add a section like this to the consumer project's CLAUDE.md:

## External Wiki: [Vault Name]

Path: /absolute/path/to/source-vault/

### Reading Order
1. Read `wiki/hot.md` first (if it exists — recent context summary)
2. Read `wiki/index.md` to find what you need
3. Read specific wiki pages as needed
4. Use grep to search `wiki/` if the index doesn't cover it

### When to Read
- When you need context about [domain] that you don't already have
- When the user asks about [specific topics]

### When NOT to Read
- [List of common tasks that don't need wiki context]
- When the user's question is clearly about code/tasks, not domain knowledge

### Token Efficiency
Don't read the entire wiki. Use the index to identify 2-3 relevant pages,
read only those. The hot cache covers recent context in ~500 words.

Step 4: Configure hot cache (if needed)

If the source vault doesn't have a hot cache but the consumer needs recency:

1. Create wiki/hot.md in the source vault
2. Configure it to be updated on every ingest:
   • ~500 words summarizing the most recent additions
   • What was just added, key findings, entities touched
3. This saves the consumer agent from crawling the full wiki for recent context

Step 5: Test the connection

Have the consumer agent try to:

1. Read the source vault's wiki/index.md
2. Find a specific topic
3. Read the relevant wiki page
4. Answer a question using that knowledge

Step 6: Log

Append to memory/log.md:

## [YYYY-MM-DD] wire | Source Vault → Consumer Project
- Source: [vault name and path]
- Consumer: [project name]
- Access type: [read-only/full/hot-cache-only]
- Topics covered: [brief list]

Real-World Example

An executive assistant project wired to a personal brain vault:

## Wiki Path: Herk Brain
Path: /Users/nate/herk-brain/

Reading order:
1. wiki/hot.md (what we just talked about)
2. wiki/index.md (find anything)
3. wiki/domain-subindex.md (deep dive)
4. Search wiki/ if needed

Don't read the wiki unless you actually need it.
You DON'T need the wiki for: scheduling, email drafts, calendar management.
You DO need the wiki for: business context, employee info, Q2 initiatives,
meeting history, strategic decisions.

Multiple Source Vaults

A consumer can reference multiple vaults:

## External Knowledge Sources

### Business Wiki
Path: /path/to/herk-brain/
Use for: business context, people, decisions

### Research Wiki
Path: /path/to/research-vault/
Use for: AI research, technical concepts, paper summaries

### Project Docs
Path: /path/to/project-vault/
Use for: architecture decisions, code concepts, implementation status

Each vault stays independent. The consumer agent reads from whichever vault is relevant to the current question.