Second Brain: Claude Code + Obsidian

Create a CLAUDE.md at the root of your vault. This tells Claude who you are, what your folder structure looks like, and how to behave.

# /your-vault/CLAUDE.md

## Who I Am
[Your identity, role, business context]

## Folder Structure (PARA)
00_Inbox/     -- quick capture
01_Projects/  -- time-bound work
02_Areas/     -- ongoing responsibilities
03_Resources/ -- reference knowledge
04_Archive/   -- completed/inactive

## How Claude Should Work With Me
- Be direct
- One step at a time when debugging
- Always explain why, not just what

Add rules that apply to specific parts of your vault. These are granular instructions Claude follows when working in each PARA area.

# /your-vault/.claude/rules/clients.md
---
path: 01_Projects/Clients/**
---
When working on client files:
- Always check Communication-Log.md first
- Update Progress.md when deliverables change
- Never delete client data without confirmation
- Use 4-file structure: README, Communication-Log,
  Technical, Progress

Create a skill that processes your inbox. This is the highest-leverage first automation. It keeps your vault clean daily with one command.

# /your-vault/.claude/skills/inbox-processor/SKILL.md

## Trigger
When user says "process inbox" or runs /inbox-processor

## Steps
1. Read every file in 00_Inbox/
2. For each file, determine category:
   - Client note → 01_Projects/Clients/Active/[match]/
   - Action item → 02_Areas/Operations/
   - Resource    → 03_Resources/[topic]/
   - Junk        → delete with confirmation
3. Extract action items
4. Move files to correct location
5. Write processing summary

Auto-memory is enabled by default. Review the settings to understand what gets stored and where. Claude saves patterns across sessions.

# Memory location for your vault:
~/.claude/projects/<vault-hash>/memory/MEMORY.md

# To view current memory:
claude "/memory"

# To manually add a memory:
claude "Remember: this client prefers French communication"

# Auto-saved entries look like:
## Client Folder Structure
Each client = 4 files: README, Communication-Log,
Technical, Progress

Open Claude Code in your vault directory and run the first command. Watch it read your CLAUDE.md, understand your structure, and start working.

# Navigate to your vault
cd /path/to/your-vault

# Start Claude Code
claude

# First command - let it orient itself
"Read CLAUDE.md and MEMORY.md. Summarize what you
know about this vault. Then process my inbox."

# Claude will:
# 1. Read CLAUDE.md - understand identity + structure
# 2. Read MEMORY.md - recall past session context
# 3. Scan 00_Inbox/ - process and route files
# 4. Save new patterns to MEMORY.md

Every morning: Claude reads 00_Inbox/, extracts action items from each note, routes to the correct PARA folder, creates tasks in Attio, archives processed items.

claude -p "Read every file in 00_Inbox/. For each file:
1. Determine category (client, internal, idea, resource, junk)
2. Extract action items with owners and deadlines
3. Move file to correct PARA folder:
   - Client notes → 01_Projects/Clients/Active/[client]/
   - Ideas        → 03_Resources/Quick_Prototypes/
   - Tasks        → 02_Areas/Operations/
   - Resources    → 03_Resources/[topic]/
4. For client actions: create Attio task via MCP
5. Write processing summary to 00_Inbox/PROCESSED.md"

Drop a meeting transcript into your vault. Claude reads it, extracts decisions and actions, updates the client's Communication-Log.md, and creates CRM tasks.

claude -p "Read the meeting transcript at 00_Inbox/meeting-[date].md.
1. Identify the client (match against 01_Projects/Clients/Active/)
2. Extract decisions made, action items (who + deadline), blockers
3. Append structured entry to Communication-Log.md
4. Update Progress.md if deliverables changed
5. Create Attio tasks for each action item
6. Move transcript to client folder"

Claude reads all files in a resource folder, finds conceptual connections, and creates an INDEX.md with [[wikilinks]]. Links appear instantly in Obsidian graph view.

claude -p "Read all files in 03_Resources/AI_Agents/.
Find connections between concepts across files.
Create INDEX.md with:
- Topic clusters grouped by theme
- [[wikilinks]] to each file
- Relationship descriptions (X depends on Y)
- Missing topics that should be documented
Use [[filename]] format so Obsidian renders graph edges."

Claude reads research notes, distills the most valuable insights, adds TL;DR sections, links to source notes, and reduces noise. Your vault gets smarter without getting bigger.

claude -p "Read all files in 03_Resources/Learning/.
For each file:
1. Add a TL;DR section (3-5 bullet points max)
2. Bold the most actionable insights in the body
3. Add [[wikilinks]] to related files in the vault
4. Score information density: HIGH / MEDIUM / LOW
5. Suggest LOW-density files for 04_Archive/
Write summary to 03_Resources/Learning/DIGEST.md"

Generate a current-state snapshot for any client. Pulls from README, Progress, Communication-Log, and Technical docs. One command for a full client briefing.

claude -p "Generate a current-state snapshot for [Client].
Read:
- 01_Projects/Clients/Active/[Client]/README.md
- 01_Projects/Clients/Active/[Client]/Progress.md
- 01_Projects/Clients/Active/[Client]/Communication-Log.md
- 01_Projects/Clients/Active/[Client]/Technical.md
Output:
- Current sprint status + blockers
- Last 3 communication entries
- Tech stack summary
- Unresolved action items"