Build Your Second Brain with Claude

Create a custom Claude skill that organizes your knowledge using PARA.

Introduction

30 Second Summary

You have ideas all day long. Script concepts in the shower, financial goals on your commute, parenting research at 2am. But without a system, those thoughts scatter across apps, sticky notes, and memory holes you will never find again.

In this project, you will build a custom Claude Skill that turns Claude into your personal knowledge management assistant. You will feed it messy brain dumps, watch it sort everything into organized notes, and then ask it to create new outputs assembled from your captured ideas.

What You'll Build

You will paste a chaotic mix of script ideas, baby care tips, and financial goals into Claude and watch it instantly separate, categorize, and format each one into a clean note. Then you will ask Claude to draft a script scene using only your captured ideas, and see it pull from your knowledge base to create something new.

By the end of this project, you'll have:

A personal capture system where you paste any raw thought and Claude instantly categorizes it using the PARA method and formats it as a saveable note.
A brain dump mode where you throw multiple mixed topics at Claude and it separates and organizes each one individually.
A creative expression workflow where you ask Claude to assemble script outlines, meal plans, or goal trackers from your previously captured knowledge.
A persistent skill that activates automatically in every new conversation, so Claude always knows your system without re-explaining it.
Secret Mission: Add a Weekly Review protocol that audits your categories and suggests what to archive or promote.

Are there any prerequisites?

You need a Claude Pro subscription (for custom skills upload) and Cursor installed on your Mac. No coding experience is required.

Before We Start

Before we dive in, let's lock in what you're about to build and why it matters to you personally.

Create Your Skill Skeleton

Claude's custom skills feature reads instructions from a specific folder structure. Without the right folder name, file name, and YAML frontmatter format, Claude treats your instructions as plain text and ignores them.

In this step, you'll create the exact folder and file skeleton that Claude expects. This gives you a valid skill that Claude can recognize and load into every conversation.

In this step, get ready to:

Create a second-brain folder on your Desktop.
Create skill.md with the required YAML frontmatter fields.
Verify the file exists from Terminal.

Create your project folder

Claude skills live in a named folder on your machine. The folder name must match the skill's name field in its configuration file.

In Cursor, open the built-in terminal by clicking Terminal in the menu bar, then New Terminal.
Move to your Desktop by running this command:

cd ~/Desktop

Create your project folder and move into it by running these commands:

mkdir second-brain
cd second-brain

Open this folder in Cursor by clicking File in the menu bar, then Open Folder.
Navigate to your Desktop, select the second-brain folder, then click Open.

You should now see second-brain as the open folder in Cursor's Explorer sidebar on the left.

Don't see the folder in the sidebar?

Make sure you selected the second-brain folder itself, not the Desktop. The Explorer sidebar should show the folder name at the top.

If your terminal closed when you opened the folder, reopen it with Terminal > New Terminal.

Add YAML frontmatter to your skill file

Every Claude skill needs a file called skill.md with YAML frontmatter at the top. Frontmatter is metadata enclosed between two --- lines that tells Claude the skill's name and when to activate it.

In Cursor's Explorer sidebar, hover over the second-brain folder name and click the New File icon (a page with a plus sign).
Type skill.md as the file name and press Enter.
Paste the following frontmatter skeleton into the file:

---
name: second-brain
description: [placeholder]
---

What does this frontmatter do?

name: second-brain sets the skill's identifier. This must match the folder name exactly.
description: [placeholder] is a temporary value. In the next step, you'll replace it with a full description that tells Claude when to activate this skill.

Save the file by pressing Cmd+S.

You should see skill.md listed in the Explorer sidebar with the content you just pasted.

File not appearing in the sidebar?

Check that you clicked the New File icon while hovering over the folder name, not above it. If the file ended up in the wrong location, delete it (right-click > Delete) and try again.

Also confirm the file is named exactly skill.md (lowercase, no spaces). Claude requires this exact filename.

Having trouble?

✔️ Awesome, I've got everything!

Great. Make sure you've saved skill.md before moving on.

ⓧ I'd like to double check the full code

Your skill.md should contain exactly this:

---
name: second-brain
description: [placeholder]
---

Verify your file from Terminal

Before you move on, confirm that your folder structure is correct by checking it from the terminal. What do you think you'll see when you list the contents of your project folder?

In Cursor's terminal, run this command:

ls ~/Desktop/second-brain/

You should see skill.md listed in the output. That confirms your folder structure is correct and Claude will be able to find your skill file.

Not seeing skill.md in the output?

If you see an empty result, the file may have been created outside the second-brain folder. Check your Desktop for a stray skill.md file.

If you see "No such file or directory", the folder itself was not created on the Desktop. Run ls ~/Desktop/ to check whether second-brain appears there.

Help me figure out why my skill.md file isn't showing up.

Your skill skeleton is in place. Next up, you'll fill it with the protocols that turn Claude into a personal knowledge management assistant.

Build Your Capture and Organize Brain

Your skill file exists, but right now it is an empty shell. Claude will read it and learn nothing useful.

A second brain that only stores information but cannot find or sort it is just a junk drawer. In this step, you will give Claude a decision framework called the PARA method so every piece of knowledge lands in exactly the right place. You will write protocols for capturing, organizing, and distilling information.

In this step, get ready to:

Write a complete skill description and PARA classification framework.
Add a Capture Protocol that formats incoming information as structured notes.
Add Organize and Distill protocols that sort and summarize your knowledge.

Write the description and PARA framework

Claude uses the description field in your frontmatter to decide when to activate your skill. A vague description means Claude might never trigger it.

In Cursor, open skill.md inside your second-brain folder.
Select all the existing content and replace it with the following:

---
name: second-brain
description: Personal knowledge management assistant using the PARA method. Helps capture, organize, distill, and express knowledge. Use when the user shares new information to save, asks to organize thoughts, wants to process or summarize notes, requests a brain dump, asks to create outputs from stored knowledge, or says weekly review.
---

# Second Brain Assistant

You are a personal knowledge management assistant. You help the user build and maintain their Second Brain using the PARA method and the CODE process (Capture, Organize, Distill, Express).

## PARA Categories

Classify all information into one of four categories based on actionability:

- Projects: Short-term efforts with a specific goal and end date (most actionable)
- Areas: Ongoing responsibilities with standards to maintain, no end date
- Resources: Topics of interest that may be useful in the future
- Archives: Inactive items from the above categories

Decision tree for classification:
1. Ask: "Is this tied to a specific goal with an end date?" If yes -> Projects
2. Ask: "Is this an ongoing responsibility requiring regular attention?" If yes -> Areas
3. Ask: "Could this be useful for reference in the future?" If yes -> Resources
4. Otherwise -> Archives

What does this code do?

The description field tells Claude both what the skill does AND when to use it. Claude reads this to decide whether to activate your skill from potentially hundreds of available skills.

The PARA framework gives Claude four buckets for any piece of information: Projects (active goals with deadlines), Areas (ongoing responsibilities), Resources (interesting reference material), and Archives (completed or inactive items). The decision tree makes classification repeatable.

Save skill.md with Cmd+S.
Scroll through the file in Cursor. You should see the --- fences around your frontmatter at the top, followed by a heading and the four PARA categories with a numbered decision tree below them.

Don't see the PARA section?

Make sure you replaced the entire file content, including the old --- fences. If you pasted below the existing frontmatter, you will have duplicate --- lines. Select all (Cmd+A) and paste again.

help me fix my skill.md file structure

Add the Capture Protocol

The Capture Protocol is the core of your second brain. It tells Claude exactly how to receive new information and format it as a structured, searchable note.

Place your cursor at the end of the file (after the last line of the PARA decision tree) and add a blank line, then paste the following:

## Capture Protocol

When the user shares new information, ideas, or content to save:

1. Acknowledge what was shared in one sentence
2. Classify it into the appropriate PARA category using the decision tree above
3. Format as a clean note using this exact structure:

```
## [Concise, Searchable Title]
Category: [Projects / Areas / Resources / Archives]
Tags: [2-4 relevant tags]
Date: [today's date]

### Key Points
- [most important point]
- [supporting detail]
- [action or next step if applicable]

### Source
[Original content, lightly cleaned up for clarity]
```

Keep key points to 3-5 bullets maximum. Prioritize actionability.

What does this code do?

This protocol gives Claude a three-step recipe: acknowledge, classify, then format. The note template ensures every capture has a searchable title, a PARA category, tags for cross-referencing, and key points distilled from the original content.

The ``` fenced block inside the protocol is a template that Claude will fill in each time you share something new.

Save skill.md.
Scroll down past the PARA section. You should see ## Capture Protocol as a new section heading, with three numbered steps and a note template inside code fences.

Seeing broken formatting?

If your code fences (```) look misaligned or the template appears outside the fences, make sure there is no extra whitespace before the opening ```. Each fence should sit on its own line with no leading spaces.

help me fix the Capture Protocol formatting in my skill.md

Add the Organize Protocol

Capturing one item at a time is useful. But what happens when you dump five ideas at once and need them sorted? The Organize Protocol tells Claude how to handle multiple items, flag ambiguous categorizations, and spot connections you might miss.

At the end of skill.md (after the last line of the Capture Protocol), add a blank line and paste the following:

## Organize Protocol

When the user asks to organize, sort, or categorize multiple items:

1. Sort all items by PARA category
2. Within each category, group by relevance or theme
3. For any item that could belong to multiple categories, state your recommendation and explain why in one sentence
4. After sorting, suggest one connection between items that the user might not have noticed

Output format:
```
### Projects
- [item]: [one-line reason it belongs here]

### Areas
- [item]: [one-line reason it belongs here]

### Resources
- [item]: [one-line reason it belongs here]

### Suggested Connection
[One sentence linking two items the user might not have connected]
```

What does this code do?

This protocol handles batch organization. Claude sorts items into PARA categories, explains borderline calls, and suggests one creative connection between unrelated items. That last step is where a second brain becomes more than a filing cabinet.

Save skill.md.
Scroll to the bottom. You should see ## Organize Protocol below your Capture Protocol, with four numbered rules and an output format template.

Organize section not appearing?

Check that you pasted after the Capture Protocol's closing line (Keep key points to 3-5 bullets maximum. Prioritize actionability.). If you accidentally pasted inside the Capture Protocol's code fence, the formatting will break.

help me check my Organize Protocol placement

Add the Distill Protocol

Storing and sorting knowledge is only half the battle. Progressive Summarization is a technique for extracting the essence of notes in layers, so you can glance at any note and instantly recall the core insight without re-reading everything.

At the end of skill.md (after the Organize Protocol's closing ```), add a blank line and paste the following:

## Distill Protocol

When the user asks to distill, summarize, or extract the essence of notes:

Apply Progressive Summarization in layers:

1. Layer 1 - Key Sentences: Bold the 2-3 most important sentences from the original
2. Layer 2 - Core Phrases: From those sentences, highlight the essential phrases
3. Layer 3 - Executive Summary: Write a 1-2 sentence summary in the user's own voice, capturing the core insight

Format:
```
### Executive Summary
[1-2 sentences in the user's voice]

### Core Insight
[The single most important phrase or idea]

### Key Passages
[2-3 bolded sentences from the original with the most essential phrases highlighted]
```

Always preserve the original content below the distillation.

What does this code do?

Progressive Summarization works in layers. First, Claude bolds the key sentences. Then it highlights the essential phrases within those sentences. Finally, it writes a one-line executive summary in your own voice.

The instruction to preserve the original content below means you never lose the full source material. You just gain faster access to the core insight.

Save skill.md.
Scroll through your entire file from top to bottom. You should see: YAML frontmatter at the top (between --- fences), then three clearly labeled protocol sections: Capture, Organize, and Distill. Each has specific instructions and a formatted template.

Missing a protocol section?

Scroll slowly through skill.md and look for three ## headings: Capture Protocol, Organize Protocol, and Distill Protocol. If one is missing, scroll to where it should be and paste it from the instructions above.

help me check if my skill.md has all three protocols

✔️ Awesome, I've got everything!

Great. Double check you have saved skill.md and move on to the next step.

ⓧ I'd like to double check the full code

Your skill.md file is longer than a quick glance, so scroll through and confirm these sections appear in order:

YAML frontmatter (--- fences) with name: second-brain and a full description field.
# Second Brain Assistant heading with an introductory paragraph.
## PARA Categories with four bullet definitions and a numbered decision tree.
## Capture Protocol with three numbered steps and a note template inside code fences.
## Organize Protocol with four numbered rules and an output format template.
## Distill Protocol with three Progressive Summarization layers and a format template.

If any section is missing or out of order, scroll up to the relevant substep and paste the code block again.

Your skill now has a brain for capturing, organizing, and distilling knowledge. Next up, you will upload it to Claude and test it with a real brain dump mixing multiple topics.

Upload and Test Your First Capture

In Step 2, you built three protocols that tell Claude exactly how to capture, organize, and distill your knowledge. Your skill.md file is fully loaded with instructions.

But that file is just sitting on your Desktop. Claude has no idea it exists. In this step, you'll upload the skill to Claude so it activates automatically in every new conversation, then test it with a real brain dump to see your second brain come alive.

In this step, get ready to:

Enable code execution in Claude's settings and compress your skill folder.
Upload your skill to Claude and toggle it on.
Test the skill by pasting a brain dump mixing three different topics.

Enable code execution and compress your folder

Claude Skills require code execution to function. You need to enable this in your settings before your skill can run.

Switch to your browser where you're signed into claude.ai.
Click your profile icon in the bottom-left corner, then click Settings.
Click Capabilities in the left sidebar.
Find Code execution and file creation and make sure the toggle is switched on.

Why does my skill need code execution?

Claude Skills use the code execution environment to load and process your skill file. Without it enabled, Claude cannot read the instructions in your skill.md even after you upload it.

Now compress your skill folder into a ZIP file that Claude can accept.

Open Finder and navigate to your Desktop.
Right-click the second-brain folder.
Click Compress "second-brain".

You should see second-brain.zip appear on your Desktop right next to the original folder.

Why a ZIP and not just the file?

Claude requires the skill to be uploaded as a ZIP containing a folder. The folder name must match the name field in your frontmatter. Your folder is called second-brain and your frontmatter says name: second-brain, so they match.

Don't see the ZIP file?

Make sure you right-clicked the second-brain folder itself, not the skill.md file inside it. The ZIP must contain the folder as its root.

Help me troubleshoot my ZIP file

Upload your skill to Claude

With your ZIP ready, you can now install the skill in Claude so it activates in every new conversation.

In Claude, click the Customize icon in the bottom-left sidebar.
Click Skills in the menu that appears.
Click the + button.
Click + Create skill.
Select Upload a skill.
Choose second-brain.zip from your Desktop and upload it.
Make sure the toggle next to your new second-brain skill is switched on.

You should see second-brain listed in your Skills panel with its toggle enabled.

Upload failing?

Check that your ZIP is named second-brain.zip and contains a folder called second-brain with skill.md inside it. The folder name must match the name field in your YAML frontmatter exactly.

Also confirm code execution is toggled on in Settings > Capabilities.

Help me fix my skill upload

Test with a real brain dump

Your skill is live. Time to throw a real test at it. You'll paste a brain dump that mixes three completely different topics and see whether Claude separates and categorizes each one using the PARA method.

Start a new conversation in Claude by clicking New chat.
Paste the following brain dump into the message box:

I had an idea for my script - what if the main character discovers their neighbor is a time traveler? Also I need to look up how to start sleep training the baby at 4 months. And I want to save $500/month toward our emergency fund.

Before you hit send, predict: how many separate notes do you think Claude will produce, and which PARA category will each one land in?

Press Enter (or click the send button) to send your message.

Claude produces 3 separate formatted notes. The script idea lands in Projects (it's tied to a specific creative goal). Baby sleep training lands in Areas (ongoing parenting responsibility). The savings goal lands in Areas (ongoing financial responsibility with a standard to maintain).

Each note has a title, category label, tags, and key points.

What just happened?

Claude read your skill instructions at the start of the conversation, recognized you were sharing information to capture, and applied the Capture Protocol. It used the PARA decision tree to classify each topic separately, then formatted each one using the exact note structure you defined.

This is the CODE process in action. You Captured raw thoughts and Claude Organized them instantly.

Claude didn't format the notes correctly?

If Claude responded with generic advice instead of formatted PARA notes, the skill may not be active. Check that the toggle is on in Customize > Skills. Also make sure you started a new conversation after uploading. Skills only activate at the start of a fresh chat.

Help me figure out why my skill isn't activating

Your second brain is alive. Claude now captures and organizes anything you throw at it, automatically sorting each thought into the right place. Next up, you'll close the loop by teaching Claude to create new things from your captured knowledge.

Complete the Loop - Create From Your Knowledge

Your second brain can capture, organize, and distill knowledge. That covers the first three stages of the CODE process. But a brain that only stores information is just a filing cabinet.

The real power is in the fourth stage: Express. This is where Claude takes your captured ideas and assembles them into something new. In this step, you'll add an Express Protocol and a Brain Dump Mode to your skill, then prove the full system works by asking Claude to write a script scene using only your captured notes.

In this step, get ready to:

Add an Express Protocol that creates outputs from your stored knowledge.
Add a Brain Dump Mode for handling streams of mixed thoughts.
Re-upload the updated skill and test the full capture-to-creation loop.

Add the Express Protocol

The Express Protocol tells Claude how to create finished outputs from your captured notes. Without it, Claude can store and organize your ideas but has no instructions for assembling them into something new.

Switch back to skill.md in Cursor.
Scroll to the bottom of the file, after the Distill Protocol section.
Add the Express Protocol by copying this code below the last line:

## Express Protocol

When the user wants to create output from their knowledge (a script, post, plan, outline, meal plan, etc.):

1. Ask what they want to create and what format they need
2. Ask them to share or reference the relevant notes and captures they want to draw from
3. Assemble the output using their captured knowledge as building blocks
4. After the output, add a "Sources" section citing which specific captures or notes each part drew from

Format the sources as:
```
### Sources
- [Section/element of output] <- drawn from [title of capture/note]
```

If the user has not shared enough material to complete the request, tell them specifically what additional notes or ideas would help.

What does this code do?

The Express Protocol gives Claude a 4-step workflow for creating new things from your stored knowledge. It asks what you want to make, gathers the relevant notes, builds the output, and then cites exactly which captures each piece drew from.

The Sources section is key. It proves Claude is using YOUR ideas, not generating generic content from its training data.

Save skill.md.

Add Brain Dump Mode

Sometimes you have a flood of unrelated ideas hitting you at once. Brain Dump Mode tells Claude how to separate those mixed thoughts into individual captures without you having to sort them first.

In skill.md, add the Brain Dump Mode section directly below the Express Protocol you just added:

## Brain Dump Mode

When the user shares a stream of mixed thoughts, multiple ideas at once, or says "brain dump":

1. Separate each distinct idea or topic (look for topic shifts)
2. Number each one
3. Classify and format each as an individual capture note using the Capture Protocol
4. End with a summary:

```
### Brain Dump Summary
Captured [X] items:
- [number] in Projects
- [number] in Areas
- [number] in Resources

Suggested next action: [one concrete next step based on what was captured]
```

What does this code do?

Brain Dump Mode handles messy, multi-topic input. When you say "brain dump" or paste a wall of mixed thoughts, Claude separates each distinct idea, numbers them, classifies each using PARA, and formats them as individual capture notes.

The summary at the end gives you a count of what landed where, plus one concrete next action so you know what to do with everything you just captured.

Save skill.md.

✔️ Awesome, I've got everything!

Great. Double check you've saved your file, then move on to the re-upload below.

ⓧ I'd like to double check the full code

Your skill.md should now have these sections in order. Scroll through and confirm each one is present:

YAML frontmatter (name and description fields).
Second Brain Assistant heading and intro paragraph.
PARA Categories (with the decision tree).
Capture Protocol (with the note format template).
Organize Protocol (with the output format).
Distill Protocol (with Progressive Summarization layers).
Express Protocol (with the Sources format).
Brain Dump Mode (with the Brain Dump Summary format).

Re-upload and test the complete system

Your updated skill file is ready. Claude is still running the old version, so you need to delete it and upload the new one.

In Claude, click Customize in the left sidebar, then click Skills.
Find second-brain in your skills list and click on it.
Click the ... button next to the toggle, then select Delete.
Click Delete in the confirmation prompt.

Now compress and re-upload the updated folder.

Open Finder and navigate to your Desktop.
If you see an old second-brain.zip file, delete it (right-click > Move to Trash).
Right-click the second-brain folder and select Compress. This creates a fresh second-brain.zip.
Back in Claude, click Customize > Skills.
Click the + button, then + Create skill, then select Upload a skill.
Upload your new second-brain.zip.
Toggle the skill on.

Why delete and re-upload?

Claude does not automatically sync changes from your local files. The version Claude uses is the one you uploaded. Deleting and re-uploading ensures Claude reads your latest instructions.

Now test the full capture-to-creation loop.

Start a new conversation in Claude.
Share 2-3 script ideas as separate messages. For example:

I have a script idea: What if the main character discovers their neighbor is a time traveler who's been subtly fixing small disasters in the neighborhood?

Wait for Claude to capture that as a formatted note, then send another:

Another script thought: The reveal scene should happen through the neighbor's garage door being left open, and the main character sees a wall of newspaper clippings about events that haven't happened yet.

Wait for Claude to capture that note too. Then send one more:

I want the tone to be warm and quirky, like a Wes Anderson film meets a neighborhood mystery. The time traveler should feel like a lovable grandparent figure.

Once Claude has captured all three, send this message:

Before you send this next message, think about what you expect Claude to produce. Will it give you generic scriptwriting advice, or something specific to the ideas you just captured?

Help me outline a script scene using the ideas I just shared.

What should I see?

Claude should produce a script scene outline that draws directly from your three specific captures. It should reference the time-traveling neighbor concept, the garage reveal scene, and the warm Wes Anderson tone.

At the bottom, you should see a Sources section listing which captures each part of the outline drew from. This proves Claude is assembling YOUR ideas, not producing generic output.

Claude giving generic advice instead of using your captures?

Make sure you started a new conversation after re-uploading the skill. Old conversations use the old skill version.

Check that the skill is toggled on in Customize > Skills. A disabled skill is invisible to Claude.

If Claude captured the ideas but produced a generic outline without a Sources section, try asking more explicitly: "Use the Express Protocol to create a scene outline from my captures above."

Help me figure out why Claude isn't using my Express Protocol.

Your second brain now handles the full loop: capture, organize, distill, and express. Every new conversation starts with Claude already knowing how to manage your knowledge and create from it.

Secret mission

Add a Weekly Review Protocol

A second brain that never gets maintained becomes a digital graveyard. Knowledge piles up, projects go stale, and connections between ideas stay hidden. Can you build a Weekly Review protocol that keeps the whole system alive and growing?

Clean Up Your Resources

Decide whether to keep your resources running, pause them to come back later, or delete them entirely. This project runs entirely locally with your existing Claude Pro subscription, so there are no ongoing costs.

Resources you used:

Local second-brain folder on your Desktop (contains skill.md).
Compressed second-brain.zip file on your Desktop.
Custom skill uploaded to Claude.ai (visible in Customize > Skills).

Keep everything running

No action needed. Choose this if you are still actively using your second brain skill and plan to keep capturing knowledge.

Your second-brain folder stays on your Desktop for future edits.
Your skill remains active in Claude and loads automatically in every new conversation.
The ZIP file stays on your Desktop in case you need to re-upload after edits.

Pause - I'll come back to this later

Disable the skill so it does not load in every conversation, but keep your files for later.

In Claude, go to Customize > Skills.
Toggle your second-brain skill off. Claude will stop loading it in new conversations.
Keep your local ~/Desktop/second-brain/ folder intact so you can re-enable or re-upload anytime.

Delete - I don't want to use this again

Remove all project resources and start fresh.

Delete the skill from Claude: go to Customize > Skills, find your second-brain skill, and delete it.
Delete the local folder and ZIP file by running this command in Terminal:

rm -rf ~/Desktop/second-brain ~/Desktop/second-brain.zip

Confirm deletion by running:

ls ~/Desktop/second-brain

You should see a "No such file or directory" message, confirming the folder is gone.

Nice Work!

You just built a personal Second Brain system powered by a custom Claude Skill. Every new conversation now starts with Claude already knowing how to capture, organize, and create from your personal knowledge.

You've learned how to:

Built a custom Claude Skill from scratch using Markdown and YAML frontmatter, turning a single skill.md file into a persistent AI behaviour layer.
Implemented the PARA method and CODE process as working software, capturing real knowledge from your life and watching Claude organize it automatically.
Proved the system works end-to-end by generating creative output assembled from your captured knowledge, with Claude citing exactly which notes it drew from.
(Secret Mission) Added an interactive Weekly Review protocol that keeps your second brain maintained and growing over time.

Ready to quiz yourself?