Making the AI-Augmented Second Brain Your Own

Last Updated on 2026-04-19 by James Croft

Part 6 in my AI-Augmented Second Brain series

Over this series, we’ve looked at a complete system for a Second Brain: a structured vault, a knowledge graph, custom AI skills, and an AI-augmented CODE workflow that transforms every stage from manual labor into a collaborative process. But I built this system for my domain, my work with my tools, my team structures, and my workflows.

That’s by design, this isn’t a product, it’s a pattern you need to adapt to your needs.

This final article covers how to make the system yours, adapting the templates, designing AI skills for your domain, connecting your data, and getting started with the companion repository.

The Obsidian AI Second Brain Repository

Everything from this series, including the folder structure, templates, base files, README, and starter AI skills, lives in a GitHub repository you can fork and customize.

Here’s what’s included:

obsidian-ai-second-brain/
├── README.md                          # CODE/PARA docs + setup guide
├── .gitignore                         # Excludes .obsidian/ workspace files
│
├── Projects/                          # Active tasks with goals and deadlines
│   └── Build My Second Brain.md       # Example: the meta-project of setting up this vault
├── Areas/                             # Ongoing knowledge domains
│   └── Building a Second Brain.md     # Example: the CODE/PARA methodology itself
├── Resources/                         # Reference material (Articles, Blueprints, etc.)
│   └── Articles/
│       └── The PARA Method - Forte Labs.md
├── Techniques/                        # Distilled actionable knowledge
│   └── Progressive Summarization.md   # Example: Forte's layered distillation technique
├── Archive/                           # Completed/inactive items
│   ├── Areas/
│   ├── Projects/
│   ├── Resources/
│   └── Techniques/
├── Journal/                           # Time-based capture (Meetings, Weekly, Monthly)
│   ├── Meetings/
│   ├── 1-1s/
│   ├── Weekly/
│   └── Monthly/
├── People/                            # Contact and colleague profiles
│
├── _Templates/                        # Note templates for each PARA type
│   ├── Template - Area.md
│   ├── Template - Technique.md
│   ├── Template - Resource.md
│   ├── Template - Project.md
│   ├── Template - Person.md
│   ├── Template - Journal Meeting.md
│   ├── Template - Journal Daily.md
│   ├── Template - Journal Weekly.md
│   └── Template - Journal Monthly.md
│
├── _Attachments/                      # Images and file attachments
│
└── .github/
    └── skills/                        # Custom AI skills
        ├── vault-knowledge-retrieval/ # Query the knowledge graph
        │   └── SKILL.md               
        ├── vault-note-creation/       # Create new notes in your Second Brain
        │   └── SKILL.md               
        └── vault-note-update/         # Updates existing/archived notes in your Second Brain
            └── SKILL.md               

What’s Included vs. What You Customize for your AI Second Brain

Included	Customize
Full PARA folder structure	Subfolder categories for your domains
9 note templates with frontmatter and callout prompts	Frontmatter properties specific to your needs
5 Obsidian Base files for dynamic indexes	Additional Base files and additional filtered queries for your workflows
3 starter AI skills (generic, no proprietary dependencies)	Additional skills for your workflows, MCP servers and tools

The included AI skills are a base for you to build on, referencing the vault’s PARA structure and template conventions, but without dependencies on any MCP servers or internal tools. They’ll for for anyone as is, and can be extended for your needs with GitHub Copilot CLI or Claude Code.

Getting Started with the Repository

1. Fork the repository from GitHub
2. Clone it locally and open it as an Obsidian vault
3. Read the README to understand the CODE/PARA structure
4. Start capturing by creating your first daily journal or meeting note, then your first domain Area note for a knowledge domain you care about
5. Try a skill by asking Copilot “what do I know about {topic}?” to invoke the vault-knowledge-retrieval skill against your vault
6. Iterate on the system to make it more valuable with every note you add and every cross-link your AI skills create

Adapting Templates for your context

The templates in the repository were designed for a knowledge worker. If that’s you, they’ll work with minimal changes. But if your domain is different, here’s how to adapt them.

Changing template frontmatter properties

Frontmatter properties should reflect what you need to query and filter. Here are some domain examples that may resonate with you:

Academic research:

---
tags: [Resource, Paper]
authors: ["Author Name"]
publication: "Journal of X"
year: 2025
last updated: 2026-03-31
---

Product management:

---
tags: [Area, Product]
team: ["Growth & Innovation"]
stage: "Discovery"  # Discovery, Build, Scale
last updated: 2026-03-31
---

Consulting project:

---
tags: [Project]
client: ["Contoso"]
engagement_type: "Advisory"  # Advisory, Implementation, Assessment
priority: 1
date_from: 2026-01-15
date_to: 2026-06-30
last updated: 2026-03-31
---

The key principles is that every frontmatter property should enable a query or filter you’ll actually use. If you won’t search by client, don’t add it. If you frequently need to find all notes for a specific author, author becomes essential.

When you change frontmatter properties, update three things:

1. The template in _Templates/ so every new notes gets the right fields
2. Any AI skill Inputs section so AI knows what properties exist and how to use them
3. And prompt AI to apply the change to any existing, active notes leveraging these templates

Changing Template section structures

Each template’s sections are guided by callout prompts. You can rename section, add new ones, or change the prompts. But keep the callout format so that both you and AI have consistent guidance on how to fill it in.

For example, if you’re in academia, you might change the Technique template’s Process section to Methodology and add a Reproducibility section:

## Methodology

> [!note] How was this research conducted?
> Describe the methods, data sources, and analytical approach in enough detail for reproducibility.

## Reproducibility

> [!warning] What's needed to reproduce these results?
> List datasets, tools, configurations, and any environment requirements.

The AI skills will adapt to section name changes.

Adapting Journals to focus areas

My daily, weekly, and monthly journal skills in my own vault have a tailored section that outline my core priorities for my role. Adding this unique context to your AI skills will ensure that the journals it captures are targeted towards you, and can evolve as you do.

Here are some example core focus areas that you can add to your AI journaling skills:

A startup founder:

### Core Focus Areas

The journal should align content to the user's established focus areas. Use these to identify and prioritize their most important work:
1. **Product Vision, Market Fit, and Differentiation** - Defining a compelling AI-driven product vision rooted in real customer problems, continuously validating product-market fit, and shaping differentiated value proposition in a rapidly evolving AI market.
2. **AI Platform, Architecture, and Data Strategy** - Designing scalable, secure, and cost-efficient AI systems (models, data pipelines, and infrastructure) while establishing a strong data strategy to ensure high-quality outputs, continuous learning, and defensible competitive edge.
3. **Execution Velocity, Automation, and GTM** - Driving rapid iteration through lean development, automation, and AI-assisted workflows, while building go-to-market engines (sales, partnerships, and growth loops) that enable efficient customer acquisition and expansion.

An academic researcher:

### Core Focus Areas

The journal should align content to the user's established focus areas. Use these to identify and prioritize their most important work:
1. **Research Direction, Problem Framing, and Scholarly Impact** - Defining novel, high-impact research questions grounded in gaps in the literature, advancing theoretical and applied knowledge, and contributing meaningful insights to the academic and broader societal landscape.
2. **Methodology, Rigor, and Experimental Design** - Designing and executing robust, reproducible studies using appropriate methodologies (quantitative, qualitative, or mixed), ensuring validity, transparency, and ethical integrity across the research lifecycle.
3. **Knowledge Dissemination, Publication, and Collaboration** - Publishing findings in high-quality venues, communicating insights clearly to academic and non-academic audiences, and collaborating across institutions and disciplines to amplify reach and impact.

A technology consultant:

### Core Focus Areas

The journal should align content to the user's established focus areas. Use these to identify and prioritize their most important work:
1. **Client Problem Framing, Strategy, and Value Realization** - Translating business challenges into clear technology strategies, aligning solutions to measurable outcomes, and ensuring initiatives deliver tangible value across efficiency, growth, and risk reduction.
2. **Solution Architecture, Integration, and Delivery Excellence** - Designing and implementing scalable, secure, and well-architected solutions that integrate across existing systems, while driving high-quality delivery through proven engineering and project practices.
3. **Stakeholder Engagement, Advisory, and Change Enablement** - Building trusted relationships with stakeholders, providing strategic guidance, and enabling successful adoption through change management, communication, and capability uplift within client organizations.

Add these in the AI skill’s Inputs to guide the AI with better capturing your daily, weekly, and monthly journals.

Designing AI skills for your domain

The four-part AI skill structure we covered in Part 4 – Role, Inputs, Instructions, and Guidelines – works for any knowledge workflow. Here’s a framework for identifying and building new AI skills.

Finding AI skills candidates

Look for workflows that are:

1. Repetitive: things that you’re doing regularly (weekly, monthly, per project)
2. Structured: processes in your day-to-day that follow a consistent pattern with identifiable steps
3. Data-rich: anywhere you can draw from multiple sources that AI can access
4. Judgment-light: the aggregation and formatting are mechanical in nature with you adding judgment to the output

Good candidates might include:

• Daily stand-up status updates based on your meeting notes and project trackers
• Client engagement summaries from backend CRM data via MCP and meeting notes
• Research literature reviews from papers captured as Resources and Area notes
• Onboarding guides for your team generated from team knowledge and process notes
• Retrospectives synthesized from sprint journals and project updates

Your AI skill design framework

For each skill candidate, answer these questions:

What inputs does the skill need?

• Which vault folders does it read from?
• Which external data sources (via MCP) does it need?
• What frontmatter properties and section structures should it expect?

Which phases does it follow?

• Discovery (scanning the vault, querying data sources)
• Analysis (correlate, deduplicate, categorize)
• Output (generate the note, report, or artifact)

What quality guidelines appy?

• What must never be fabricated?
• What must always be cited?
• What scoped constraints prevent the skill from wandering?

These answers give you your skill’s Inputs, Instructions, and Guidelines.

Connecting your data sources via MCP

The Model Context Protocol is what lets skills reach beyond the vault. In my system, WorkIQ bridges M365 activity and Azure DevOps provides work items. Your system may have different data sources.

What MCP enables for your AI Second Brain

MCP servers expose external tools that skills can invoke during their workflow. The pattern is always the same:

• The skill instruction says when to query the external source (“After gathering vault notes, query {tool} for additional context”)
• The MCP server handles the how, including authentication, data retrieval, response formatting
• The skill instruction says what to do with the results (“Synthesize into the relevant section, prioritizing user input over external data”)

Common MCP Integration Patterns for your AI Second Brain

Source	MCP Server	Skill Use Case
Microsoft 365 / Google Workspace	Microsoft Work IQ CLI / Google Workspace CLI	Enrich journals from meetings and emails
Azure DevOps / GitHub / Jira	Azure DevOps MCP / GitHub MCP / Atlassian MCP	Enrich details in journals and Projects from work items
Microsoft Teams / Slack	Microsoft Work IQ CLI / Slack MCP	Enrich meeting journals and people notes with chat history
Salesforce / HubSpot	Salesforce DX MCP / HubSpot MCP	Enrich client notes with engagement history

The MCP ecosystem is growing rapidly, and I highly recommend checking the MCP server registry for any existing servers before considering to build your own. Always look for officially supported MCP servers from your service providers.

Maintaining and Evolving your AI Second Brain

A knowledge system is a living thing. Here’s how to keep it healthy.

Periodic Reviews

Weekly: Review your journals. The AI’s synthesis is a starting point, add your own reflections, correct any misinterpretations, and ensure the tone matches your voice.

Monthly: Scan for Areas with no associated Techniques (gap detection from Part 3). These represent knowledge domains where you’ve captured but haven’t distilled. Pick one and write a Technique, or ask AI to help you draft one from the Area’s content and related Resources using skills.

Quarterly: Review Archive for notes that have become relevant again. Everything can feel like it’s moving fast right now, and an archived Area might deserve reactivation. Run a knowledge retrieval query on topics you’ve been working on and see if the results surface anything from Archive that should come back.

AI Skill Iteration

Your skills will need refinement. The pattern:

1. Use the skill for its intended task
2. Evaluate the output. Did it miss something? Include something irrelevant? Format incorrectly?
3. Trace the issue to a specific instruction step or missing guideline
4. Add or refine that instruction or guideline
5. Test again with a different query to verify the fix doesn’t break other cases

Most improvements come from adding guidelines, not restructuring instructions.

Handling your Second Brain growth

As your vault grows to hundreds or thousands of notes, the knowledge base scan becomes more valuable. The larger the vault, the more likely a new note has related content somewhere that should be cross-linked.

The templates and base queries scale naturally because they’re driven by the graph structure. The system doesn’t require more maintenance as it grows, it requires the same maintenance applied to more content.

However, If search performance becomes an issue, consider organizing subfolders more granularly and using more specific glob patterns in your skill instructions.

What’s next for the Self-Improving AI Second Brain?

Here’s something remarkable about this approach: the skills you write today will produce better results tomorrow without changes.

As AI models advance in reasoning, context handling, and instruction following, the same SKILL.md files will generate higher-quality output. The models improve; your grounding (the vault’s structure, templates, and graph) remains the constant. You don’t always need to rewrite skills when a new model ships. The grounding is what makes the output reliable, and the model is what makes it intelligent.

And as your vault grows, with more Areas filled in, more Techniques distilled, more cross-links connecting ideas, the knowledge base scan produces richer results. AI discovers more context, makes more connections, and generates more complete output. The content improves the AI’s effectiveness, and the AI’s effectiveness improves the content.

This is the compounding flywheel we introduced right at the beginning of this series: structure enables AI, AI strengthens the graph, the graph guides the AI. It’s not a one-time setup. It’s a system that gets better the more you invest in it.

The AI-Augmented Second Brain Series in Review

Let’s close by tracing the full arc:

1. Why Your Second Brain Needs an AI Companion: The problem (knowledge management doesn’t scale manually) and the thesis (three pillars: structure, graph, skills)
2. Designing a Machine-Readable Knowledge Base with Obsidian: The foundation: PARA folders, templates with frontmatter, base files as dynamic indexes, and the README as dual-purpose documentation
3. Building out your Knowledge Graph in Markdown: The connections: wikilinks as graph edges, backlinks as implicit relationships, base queries as dynamic tables, and gap detection through structure
4. Your First AI Skill for Knowledge Management: The intelligence: what AI skills are, the four-part instruction pattern, the knowledge base scan, and building a new skill from scratch
5. Leveraging the AI-Augmented CODE Workflow: The workflow: AI-Orchestrated Capture, AI-Augmented Organize, Human & AI Distill, and AI-Supported Express
6. Making It Your Own: The adaptation: customizing templates, designing domain skills, connecting data sources via MCP, and getting started with the companion repository

The methodology of CODE and PARA remain the workflow and the structure. What’s changed is the capacity, the ability to maintain, distill, and express knowledge at a scale that manual effort simply can’t match.

Your Second Brain is no longer a filing cabinet. It’s an active collaborator in its own growth. And now it’s yours to build.

Found this article useful?

Thank you for taking the time to read this article. I’m committing to sharing more of my knowledge through my blog and open-source projects! You’ll also catch me in casual conversation on Bluesky and LinkedIn too.

If you enjoy what you see, please consider subscribing to get a notification when new articles go live!

---

This article is part of my AI-Augmented Second Brain series.

You can find the companion template repository for an Obsidian vault at obsidian-ai-second-brain on GitHub.