Results Studio

A starter vault for getting real work done with AI agents.

A structured personal knowledge vault designed to work with any AI coding agent that reads AGENTS.md (Claude Code, Cursor, Windsurf, Aider, Cline, and others). Based on the PARA method (Projects, Areas, Resources, Archive) with agent-facing context files baked in.

This is a personal share, not a product. It is a folder structure, a few conventions, and a set of context files that make AI agents useful across sessions. No roadmap, no support, no guarantees. If it's useful, take it and adapt it freely.

Privacy first. Once you start using this vault you will accumulate client names, pipeline notes, meeting context, and personal details. Default to a private repository. Do not push to a public repo unless you have scrubbed it completely.

---

Why this works

AI agents are stateless by default. Every time you open a new session, the agent has no memory of what you're working on, who your clients are, or what decisions you've already made. You end up re-explaining context constantly.

This vault solves that by keeping machine-readable context files (AGENTS.md, HUMAN.md) next to the work they describe. The agent reads them at session start and picks up where you left off -- without you doing the re-briefing.

---

What you get

• A folder hierarchy that works for consulting, advisory, or general knowledge work
• AGENTS.md files that brief AI agents at the start of each session
• HUMAN.md files that keep you oriented between sessions
• A root HUMAN.md dashboard that rolls up active work across all projects
• A daily journal (00-Notebook/) with an Obsidian-ready config -- optional, but pre-wired
• Example fictional projects, areas, and journal entries to show the pattern in action
• A root AGENTS.md that wires everything together for any AI agent

---

Quick start

1. Get the files

Option A -- Fork on GitHub

gh repo fork scottheffield/results-studio my-vault
cd my-vault

Option B -- Clone directly

git clone https://github.com/scottheffield/results-studio.git my-vault
cd my-vault

Option C -- Download ZIP

Download and unzip to a folder of your choice. No git required to start.

---

2. Put it somewhere you can access from any device

The vault is just files. Put the folder anywhere:

Storage	Notes
Google Drive	Use "Drive for Desktop" to sync a local folder. Works with any file editor.
OneDrive	Sync client keeps a local copy. Full path access for agents.
Dropbox	Same pattern -- local sync folder.
Local only	Fine if you work from one machine.

Pick a path you can remember. Example: ~/Documents/my-vault or G:/My Drive/Vault.

---

3. Clear out the demo content

The vault ships with fictional projects, areas, and a sample archive so you can see the pattern in action. Once you've had a look, replace it with your own.

The easy way: open the vault in Claude Code (or Cursor) and paste this prompt:

Reset this vault for my own use. Do the following:

1. Delete every folder inside 02-projects/, 03-areas/, 05-archive/, and 01-inbox/ except 01-inbox/AGENTS.md.
2. Delete the demo journal under 00-Notebook/Work Journal/ -- but keep 00-Notebook/README.md, 00-Notebook/AGENTS.md, and 00-Notebook/Templates/.
3. Empty HUMAN.md at the repo root, keeping only the section headers (## Due soon, ## Active projects, ## Business development, ## Learning and development, ## Personal).
4. Replace every file in 04-resources/About-Me/ with a blank version using the template instructions at the top of each file.
5. Leave the root AGENTS.md, CLAUDE.md, README.md, LICENSE, .gitignore, .obsidian/, and 04-resources/Templates/ untouched.
6. Stop and confirm before deleting anything.

The manual way: delete the sample folders under 02-projects/, 03-areas/, 05-archive/, and any files under 01-inbox/ except its AGENTS.md. Delete 00-Notebook/Work Journal/2026-04/ (keep the rest of 00-Notebook/). Empty out HUMAN.md and 04-resources/About-Me/AboutMe.md. Keep the templates.

---

4. Add your first real project

Once the demo content is cleared, create a project folder for real work:

1. Create a folder under 02-projects/ with a short hyphenated name (02-projects/Client-Name-Initiative/)
2. Copy 04-resources/Templates/project-AGENTS-template.md into it and rename it AGENTS.md
3. Copy 04-resources/Templates/project-HUMAN-template.md into it and rename it HUMAN.md
4. Fill in the blanks -- who, what, deliverable, next steps

Or ask your agent to do it:

Create a new project folder for [project name]. Copy the templates from 04-resources/Templates/, fill in what you know from our conversation, and flag anything that needs my input.

---

5. (Optional) Initialize git

Git gives you version history and a backup. Recommended if you're technical.

If you cloned or forked, git is already initialized -- skip to the commit step.

If you downloaded a ZIP:

cd my-vault
git init
git add .
git commit -m "init: vault from template"

# push to a PRIVATE repo -- this vault will contain personal and client context
gh repo create my-vault --private --source=. --push

Important: Default to a private repo. Your vault will accumulate client names, pipeline notes, and personal context over time. Don't push it to a public repository.

Note: If your vault lives in a cloud-synced folder (Google Drive, OneDrive, Dropbox), you already have a backup mechanism. Git is still useful for history and diffs.

---

6. Personalize 04-resources/About-Me/AboutMe.md

This file is what agents read to understand who you are and how to work with you. Replace the placeholder content with your own context. The more specific you are, the better the agent performs.

---

7. Open with your AI agent

Open the vault folder in whatever AI coding agent you use. Most modern agents read AGENTS.md automatically at session start. A few examples:

• Claude Code: cd my-vault && claude. Also reads CLAUDE.md (shipped as a pointer to AGENTS.md).
• Cursor: open the folder as a workspace -- reads AGENTS.md natively.
• Windsurf, Aider, Cline, and others: same pattern -- open the folder, the agent picks up AGENTS.md.

Regardless of tool, a good opening prompt is:

"Read HUMAN.md and brief me on what's active."

---

Folder structure

my-vault/
├── AGENTS.md                  ← AI agent master instructions
├── HUMAN.md                   ← Your active dashboard (read this first each session)
├── LICENSE
├── README.md
├── 00-Notebook/               ← Daily work journal (Obsidian-ready, optional)
├── 01-inbox/                  ← Unprocessed captures -- triage before moving
├── 02-projects/               ← Work with a finish line (one subfolder per project)
├── 03-areas/                  ← Ongoing responsibilities (no end date)
├── 04-resources/              ← Reference material and context files
└── 05-archive/                ← Inactive projects and historical material

Each project and area subfolder contains:

• AGENTS.md -- what the AI agent needs to know about this folder
• HUMAN.md -- what you need to know: next steps, key people, decisions

---

Daily journal (optional Obsidian)

00-Notebook/ is the daily work journal -- a place for the daily focus, tasks, scratch notes, and small captures. It complements 02-projects/ and other PARA folders rather than replacing them. Finished artifacts still graduate to the project folder; the journal is where the day's thinking happens.

The vault is pre-wired for Obsidian:

• .obsidian/daily-notes.json sets the path to 00-Notebook/Work Journal and points at the template in 00-Notebook/Templates/Daily Note.md.
• Install Obsidian, open the vault folder as a vault, and the Daily Note ribbon creates today's note in Work Journal/YYYY-MM/YYYY-MM-DD/YYYY-MM-DD.md with no extra setup.

Obsidian is optional. The journal is just Markdown -- any editor works. The pre-wired config saves the click for anyone who does use Obsidian.

Capture convention. Inside any day note, lines starting with - Idea: are designed to be harvested by a script of your choice (Node, Python, bash, n8n, anything). The template ships no harvest script -- pick the runtime that fits your stack. Full details in 00-Notebook/README.md and 00-Notebook/AGENTS.md.

The demo journal under Work Journal/2026-04/ shows the pattern in use across a few fictional days.

---

Session ritual

Starting a session:

1. Agent reads HUMAN.md (root) -- surfaces what's active and urgent
2. Agent checks 01-inbox -- asks you what to do with anything there
3. You tell the agent which project or area you're working in
4. Agent reads that folder's AGENTS.md

Ending a session:

1. Agent updates HUMAN.md (root) with any changes to project status
2. Agent updates the folder-level HUMAN.md for the area you worked in
3. Commit changes if using git

See EXAMPLE-SESSION.md for what a session actually looks like.

---

What this is not

• Not a replacement for Notion or your notes app -- it is a folder structure, not a tool
• Not a task manager (use whatever you use now)
• Not a database or CRM
• Not a file-syncing tool (use Drive, OneDrive, or Dropbox for that)

Obsidian is not required, but the vault is pre-wired for it. See the Daily journal (optional Obsidian) section above.

---

Tips

• Keep AGENTS.md files lean. A tight, current brief beats an exhaustive one. If the file grows past a page or two, something should probably move into a linked note.
• HUMAN.md is for you, not the agent. Write it in shorthand if that helps you -- the agent will still read it fine, and you'll actually keep it updated.
• Don't over-engineer the root dashboard. It's a markdown rollup. If you find yourself building logic around it, step back.
• One project per finish line. If a project folder has three unrelated deliverables in it, split it.
• Archive aggressively. Finished projects move to 05-archive/ with a short summary. Don't let 02-projects/ turn into a graveyard.
• Tool-specific context files. AGENTS.md is the canonical file. A one-line CLAUDE.md ships alongside it pointing Claude Code to the same content. If your agent uses a different filename convention, add a similar pointer -- don't duplicate the instructions.
• Don't edit root HUMAN.md while an agent is actively writing to it. In long sessions with parallel agents, you can end up with conflicting writes. One writer at a time.

---

Going further

Once the basics are in place, there are a few directions worth exploring on your own:

• Skills and slash commands. Most agents support reusable, named capabilities (skills, commands, rules, recipes). They pair well with a vault -- a single /triage-inbox or /end-of-session command can replace a long prompt. skills.sh is one place to see what's out there and learn the pattern.
• Automation scripts. A .scripts/ folder at the repo root is a natural home for vault-level automation (dashboard regeneration, link checking, weekly digest generation). Keep it lean.
• A personalized "About Me" layer. 04-resources/About-Me/ can hold more than just a bio: tool preferences, positioning, behavioral style, business context. The more the agent knows about you, the less re-briefing you do.
• Tailored rollup buckets. The default buckets in root HUMAN.md are a starting point. Rename them to match how you actually think about your work.

This vault gestures at what's possible. Extending it is your job, and the fun part.

---

Customization

• Add or rename 02-projects/ subfolders freely -- one per engagement, initiative, or deliverable
• 04-resources/Templates/ has starter AGENTS.md and HUMAN.md templates for new projects
• Modify the root AGENTS.md to adjust agent behavior: tone, session rituals, decision filters, rollup bucket names

---

Attribution

Template structure and conventions by Scott Heffield of Results Oriented LLC. Based on the PARA method by Tiago Forte.

MIT License -- fork, adapt, and share freely.