Customizing Your Setup

Make your brain work the way you work. Start with defaults, customize what bugs you.

Customization Levels

From easiest to most advanced:

Level 1: Your Research Sources (Easy - 10 minutes) Add newsletters and content sources you care about

Level 2: Your Preferences (Medium - 20 minutes) Personalize how Claude works with you

Level 3: Custom Skills (Advanced - 1-2 hours) Create auto-activating expertise

Level 4: Custom Agents (Expert - 2-4 hours) Build specialist workers for your workflows

Start at level 1. Only go deeper when you need it.

Level 1: Your Research Sources

Add your own newsletters and content:

Add a Newsletter Source

# Create folder for new source
mkdir -p research/newsletters/your-source

# Create index file
touch context/ideas/your-source.md

# Add header
echo "# Your Source Name

Summary insights from Your Source newsletter.

---

" > context/ideas/your-source.md

Create Gmail Filter

Open Gmail
Search for: from:newsletter@yoursource.com
Click "Create filter"
Check "Apply label" → Choose "8020brain"
Check "Also apply to matching conversations" (backfills past emails)
Create filter

Test It

"Check newsletters"

# Verify new source appears
ls research/newsletters/your-source/

Done. Your newsletter now feeds your brain automatically.

Add a YouTube Channel

Save videos you want to remember:

# Create channel folder if needed
mkdir -p research/youtube/your-channel

# Fetch a video
/fetch https://youtube.com/watch?v=xyz123

# Creates transcript + summary automatically

Each video becomes a searchable document in your brain.

Add a Podcast (via Transcript)

Convert audio to searchable text:

# Convert podcast to transcript using Whisper or similar
# Save transcript as markdown

# Create podcast folder
mkdir -p research/podcasts/your-podcast

# Save transcript
# (paste or save your transcript.md here)

# Create index
touch context/ideas/your-podcast.md

Now brain-advisor can search podcast content too.

Remove a Source

Not using a source anymore?

# Remove Gmail filter (in Gmail UI)
# Archive or delete the folder
mv research/newsletters/old-source ~/archive/

# Remove index
rm context/ideas/old-source.md

Your brain only contains what you actively use.

Level 2: Your Preferences

Personalize how Claude Code works with you.

Edit Your Global Config

# Open your global preferences
open ~/.claude/CLAUDE.md

What to customize:

About You

## About Me
- **Name:** [Your Name]
- **Company:** [Your Company]
- **Role:** [Your Role]
- **Location:** [City, Country]

Claude uses this context in responses.

Your Focus Areas

## My Focus Areas

**Topics I care about:**
- AI automation for agencies
- Google Ads optimization
- Content marketing strategy

**Decision frameworks I use:**
- 80/20 principle
- Jobs To Be Done
- First principles thinking

**Tools I use daily:**
- Claude Code
- Google Ads
- Notion

Helps Claude understand your context and priorities.

Communication Preferences

## Communication Preferences

**Tone:** Professional but conversational
**Format:** Concise, action-oriented
**Examples:** Always show real examples
**Emojis:** Only when explicitly requested

Claude adapts to your style.

Your Industry Context

## Industry Context

**Industry:** Digital marketing agency
**Clients:** B2B SaaS companies
**Size:** 5-50 employees
**Budget range:** $10k-100k/month ad spend

Makes responses more relevant to your world.

Project-Specific Preferences

Working on a specific project with different needs?

Edit project-level config:

# Open project config
open ~/Projects/your-project/CLAUDE.md

Example use cases:

Client project with specific requirements
Personal project with different tone
Experiment with specific constraints

Project config overrides global config.

Brain-Specific Preferences

Want to customize just your brain behavior?

# Open brain config (from within your brain folder)
open CLAUDE.md

Common customizations:

Which research sources to prioritize
How to format responses
What to include/exclude in searches

Level 3: Custom Skills

Create auto-activating expertise for your workflows.

When to Create a Skill

Create a skill when you have a workflow that:

You do repeatedly
Has clear trigger words
Benefits from automatic activation
Follows consistent steps

Examples:

"Analyze client performance" triggers Google Ads analysis
"Draft proposal" triggers proposal template and research
"Weekly review" triggers summary of recent research

Create a Simple Skill

# Create skill folder
mkdir -p .claude/skills/my-skill

# Create skill definition
touch .claude/skills/my-skill/skill.md

Edit skill.md:

# My Custom Skill

**Auto-activates when:** User mentions [trigger words/patterns]

**Purpose:** [What this skill does in one sentence]

**Tools allowed:** Read, Write, Grep, Glob

---

## Instructions

When this skill activates:

1. [First step]
2. [Second step]
3. [Third step]
4. [Return result to user]

---

## Examples

**Example 1:**

User: "[trigger phrase]"

Action:
- [What the skill does]
- [Expected output]

**Example 2:**

User: "[another trigger]"

Action:
- [What the skill does]
- [Expected output]

---

## Notes

- [Important considerations]
- [Edge cases to handle]
- [When NOT to activate]

Test Your Skill

# Try your trigger phrase
"[your trigger words]"

# Verify skill activates
# Check result matches expectations

If it doesn't activate:

Check trigger words are clear
Make sure purpose is specific
Test with exact phrases from examples

Skill Template (Copy This)

# [Skill Name]

**Auto-activates when:** User says/mentions [specific trigger]

**Purpose:** [One clear sentence about what this does]

**Tools allowed:** [List of tools: Read, Write, Bash, etc.]

---

## Instructions

When activated:

1. Understand user intent: [What are they asking for?]
2. Gather information: [What data do you need?]
3. Process: [What do you do with the data?]
4. Respond: [What do you return to user?]

Be specific. Give Claude exact steps.

---

## Examples

Provide 2-3 clear examples of:
- User input
- What skill should do
- Expected output

---

## Edge Cases

When NOT to activate:
- [Condition 1]
- [Condition 2]

When to ask for clarification:
- [Scenario 1]
- [Scenario 2]

Level 4: Custom Agents

Build specialist workers for complex workflows.

When to Create an Agent

Create an agent when you have a task that:

Requires multiple steps
Needs focused attention
Benefits from isolated context
You or Claude invoke explicitly

Examples:

Competitive analysis agent (research competitors)
Content audit agent (analyze all content)
ROI calculator agent (calculate and report ROI)

Create a Simple Agent

# Create agent folder
mkdir -p .claude/agents/my-agent

# Create agent definition
touch .claude/agents/my-agent/agent.md

Edit agent.md:

# My Custom Agent

**Purpose:** [What this agent specializes in]

**Invoked by:** User or Claude when [specific condition]

**Tools available:** [List tools this agent can use]

---

## Agent Instructions

You are a specialist in [domain].

When invoked, your job is to:

1. [Specific task 1]
2. [Specific task 2]
3. [Specific task 3]
4. Return [specific output format]

---

## Process

**Step 1: [Name]**
[Detailed instructions]

**Step 2: [Name]**
[Detailed instructions]

**Step 3: [Name]**
[Detailed instructions]

---

## Output Format

Return results as:

[Describe exact format]
[Include example if helpful]

---

## Examples

**Example 1:**

Input: [What user provides]
Process: [What agent does]
Output: [What agent returns]

**Example 2:**

Input: [What user provides]
Process: [What agent does]
Output: [What agent returns]

Invoke Your Agent

# Explicit invocation
"Use my-agent to [task description]"

# Or let Claude recommend it
"I need to [task that matches agent's purpose]"

Agent Template (Copy This)

# [Agent Name] Agent

**Purpose:** Specialist in [specific domain/task]

**Invoked when:** [Explicit user request OR specific scenario]

**Tools:** [List available tools]

---

## Core Competency

You are an expert at [specific skill/domain].

Your goal: [Clear objective]

Your approach: [High-level methodology]

---

## Process

### 1. Understand the Request
- [What information do you need?]
- [What questions should you ask?]

### 2. Gather Information
- [Where do you look?]
- [What do you read?]

### 3. Analyze
- [How do you process the information?]
- [What are you looking for?]

### 4. Report
- [What format?]
- [What details to include?]

---

## Examples

[Provide 2-3 detailed examples of the agent in action]

---

## Quality Checks

Before returning results:
- [ ] [Checklist item 1]
- [ ] [Checklist item 2]
- [ ] [Checklist item 3]

File Organization

Prefer different folder structure? Adapt these:

your-brain/
├── research/          # Your research content
│   ├── newsletters/   # Newsletter summaries
│   ├── youtube/       # Video transcripts
│   ├── podcasts/      # Podcast transcripts
│   └── articles/      # Saved articles
│
├── context/           # Indexed knowledge
│   ├── ideas/         # Summary indexes by source
│   └── business/      # Business documentation
│
├── .claude/           # Brain configuration
│   ├── memory/        # Runtime facts
│   ├── skills/        # Auto-activating expertise
│   ├── agents/        # Specialist workers
│   └── commands/      # Manual workflows
│
└── !inbox/            # Quick capture

Want different names?

Update your CLAUDE.md to reference your structure:

## My File Organization

- Research lives in: `knowledge/`
- Indexes live in: `summaries/`
- Capture goes to: `inbox/`

[Adjust instructions to use your paths]

Your brain, your structure.

Pro Tips

Start with Defaults

Don't customize until something bugs you.

The defaults work for 90% of use cases. Customize the 10% that matters to you.

Copy Existing Examples

Don't start from scratch.

# Copy an existing skill as template
cp -r .claude/skills/existing-skill .claude/skills/my-new-skill

# Edit to match your needs

Faster and fewer mistakes.

Test Incrementally

Don't build a complex skill/agent all at once.

Start with basic version
Test it
Add one feature
Test again
Repeat

Build confidence before adding complexity.

Document Your Customizations

Keep notes on what you changed and why:

# Create customization log
touch .claude/CUSTOMIZATIONS.md

Track:

What you customized
Why you made the change
When you did it
How to revert if needed

Future you will thank present you.

Share Useful Patterns

Created something valuable? Share in Circle.

Others might benefit, and you might get improvement ideas.

Community-driven customization is powerful.

What Can Go Wrong

Skill Not Activating

Causes:

Trigger words too vague
Conflicts with other skills
Syntax error in skill file

Fixes:

Use more specific triggers
Check no other skill has same triggers
Validate markdown syntax
Test with exact example phrases

Agent Runs But Gives Wrong Results

Causes:

Instructions unclear
Missing key steps
Wrong tools available

Fixes:

Add more specific instructions
Break down steps further
Grant additional tools if needed
Add quality checklist at end

Customization Conflicts

Causes:

Global and project configs conflict
Multiple skills with same triggers
Memory hierarchy confusion

Fixes:

Remember: Inner overrides outer
Review all CLAUDE.md files in hierarchy
Rename conflicting triggers
Be explicit about which config applies when

Examples of Good Customizations

Custom Research Source

Added Substack writer:

mkdir -p research/newsletters/my-expert
echo "# My Expert\n\n" > context/ideas/my-expert.md
# Created Gmail filter
# Now auto-captures their insights

Client Project Preferences

Different tone for client work:

## Client Project: Acme Corp

**Tone:** Formal and detailed
**Format:** Include rationale for all recommendations
**Focus:** ROI and business impact
**Always mention:** Industry benchmarks

Weekly Review Skill

Auto-activates on "weekly review":

# Weekly Review Skill

**Triggers:** "weekly review", "week in review"

**Actions:**
1. List recent newsletters captured
2. Show key themes from indexes
3. Highlight notable quotes
4. Suggest follow-up research

Competitive Analysis Agent

Researches competitors:

# Competitive Analysis Agent

**Purpose:** Research competitor positioning and strategy

**Process:**
1. Search research for competitor mentions
2. Compile quotes and insights
3. Identify patterns
4. Create comparison summary

Next: See how others use their brains →

Quick Reference

Customization levels:

Research sources (easy)
Preferences (medium)
Custom skills (advanced)
Custom agents (expert)

Add newsletter source:

mkdir -p research/newsletters/source-name
touch context/ideas/source-name.md
# Create Gmail filter

Edit preferences:

open ~/.claude/CLAUDE.md        # Global
open ~/Projects/CLAUDE.md       # Projects (if you use this level)
# Brain-specific: open CLAUDE.md from within your brain folder

Create skill:

mkdir -p .claude/skills/skill-name
# Copy template, customize, test

Create agent:

mkdir -p .claude/agents/agent-name
# Copy template, customize, test

Pro tips:

Start with defaults
Copy existing examples
Test incrementally
Document changes
Share useful patterns