Your Second Brain
starts with one folder

One folder. Your files. One prompt. Primed for work.

Your Second Brain is a folder of context your AI wakes up knowing. Briefs, decisions, working memory, the people, the project state — wired in once, loaded in 60 seconds.

Built for IT Pros — business analysts, architects, testers, DBAs, support engineers. Runs three ways:

  • On your computer — the AI reads a folder on your PC or Mac. Best for: Claude Cowork, Claude Code, Cursor, Windsurf, Obsidian.
  • In a cloud Project — your files live in a personalized AI space the assistant remembers across chats. Best for: Claude Projects (claude.ai), ChatGPT Projects, Gemini, Microsoft 365 Copilot.
  • In any chat — paste the prompt, drop your files in. No setup, no account upgrades. Works on the free tier of claude.ai or chatgpt.com.

60-second daily load. 10-minute weekly reset. That's the whole practice.

// step 01

Make a folder

Anywhere on your computer. Call it whatever.

// step 02

Drop your stuff in

Briefs, meeting notes, decisions, schemas — whatever you already work from.

// step 03

Run the prompt

Your AI reads the folder, asks a few questions, builds the structure. Done.

// Downloads printdark
Module 01 only Personal Knowledge Database PDF PDF
Setup Companion only Three paths · six tool walkthroughs PDF PDF
All-in-one Module 01 + Setup Companion, single PDF PDF PDF
// Your Second Brain

Three files. Built in 10 minutes by the prompt below. Updated in 60 seconds a day. Everything else is optional bonus.

// 01

identity.md

rarely changes

Who you are. How you want AI to answer. Your non-negotiables.

// 02

project-name.md

weekly

One brief per active project. Goals, status, decisions, stakeholders, deadlines.

// 03

state.md

daily

What you touched today. What's blocked. What's next. This week's working memory.

+ Wrappers (Path A only) — CLAUDE.md and AGENTS.md at the project root tell filesystem agents to read the three files above on every session. Skip if you're on Path B or C.
+ Patterns folder — patterns/ holds prompt templates and rubrics you've found work for you. Add over time.
+ Artifacts — code, docs, tickets linked from your project file. Never copy-pasted — the moment you copy, it goes stale.
// Step 01 · Pick your path

Where do you actually work with AI?

A

On your computer

For people whose work runs in an IDE, terminal, or local folder. The agent reads your folder directly. Setup once, future sessions auto-load via the CLAUDE.md / AGENTS.md wrappers.

Claude CodeCursorWindsurfCodex CLIAiderCopilot AgentGemini CLI
// Path A walkthrough Print PDF Dark PDF
B

In a cloud Project

For people who live on the provider's website. Upload your three files to a Project once; the assistant remembers them across chats. Best balance of "no install" and "remembers your context".

Claude ProjectsChatGPT ProjectsGemini GemsCustom GPTsNotebookLMM365 Copilot
// Path B walkthrough Print PDF Dark PDF
C

In any chat

For people just trying it out, or who jump between tools. Attach files at session start, paste the prompt, save outputs. No setup, free tier works — you'll re-attach next session.

claude.aiChatGPTGemini appPerplexity
// Path C walkthrough Print PDF Dark PDF

Click-by-click setup for each path → Setup Companion PDF in the Downloads block above.

// Step 02 · Gather your context

You don't need to write anything new — just point your AI at 3–5 things you already work from. Pick from any of these:

Project context

  • Brief, spec, or PRD
  • Roadmap or strategy doc
  • Slide deck for the work
  • Stakeholder meeting notes
  • Sprint retro or post-mortem

Recent decisions & state

  • Slack / Teams chat exports
  • Email threads where decisions got made
  • Last week's status update

People & process

  • Stakeholder list — who cares, who approves
  • Tickets (Jira / Linear / Asana)
  • SLA or scope documents

Technical shape (when relevant)

  • System diagrams, architecture sketches
  • DB schema or data dictionary
  • API contracts
  • Runbook or ops playbook

You don't need all of this. Pick the 3–5 documents that, if your AI had them open, would stop you having to explain the project from scratch. Drop them into your folder (Path A), upload to your Project (Path B), or attach in chat (Path C).

// Step 03 · Run the prompt

Pick the prompt that matches the path you chose. Copy it, paste into your AI, answer 3–5 short questions, save what it gives you. ~10 minutes.

A · On your computer
› Show the Path A prompt // filesystem write · wrappers included
context-stack-builder · path A.md 
Act as my Context Stack builder.

## Safety contract (Path A — filesystem write)

- Read only from the folder I've given you access to. No traversal beyond.
- Write only inside `context-stack/` and the wrapper files `CLAUDE.md` / `AGENTS.md` at the project root. Nothing else. Never modify or delete anything outside that scope.
- Skip credentials (keys, passwords, tokens, `.env` files) and anything marked confidential. If unsure, skip and note.
- Log every file you read and every file you write in the debrief.

## What to build

Three files in `context-stack/`:

1. **`identity.md`** — who I am, how I want you to answer, my non-negotiables.
2. **`project-<slug>.md`** — this project's brief (goals, status, decisions, stakeholders, deadline). Lowercase slug, e.g. `project-feedrunner.md`.
3. **`state.md`** — this week's working memory.

Plus two auto-loader wrappers at the project root:

4. **`CLAUDE.md`** — read by Claude Code on every session.
5. **`AGENTS.md`** — read by Codex CLI, Aider, and the cross-tool convention.

Both wrappers contain the same short content (template below).

## Steps

### 1. Inventory silently
Scan the folder. Don't tell me what you found yet. Priority:
- High: charters, briefs, PRDs, scope docs, meeting notes, decisions.
- Medium: PDFs, DOCX, PPTX (data-extract), exported chats.
- Low: filenames as weak signal.

Skip secrets.

### 2. Ask 3–5 short questions
Batch in one message. Skip what the inventory already answered. Typical:
1. My role.
2. How I want you to answer (direct/detailed/sparse, prose/bullets, push back or default-agree).
3. The one thing blocking or driving this week.
4. The last significant decision.
5. Who else is involved.

### 3. Draft the three files
Use the templates below. Rules:
- Specific beats complete. 4 specific lines beats 20 generic.
- `[TBD: short reason]` when signal is weak. Don't invent.
- Recent-first in `state.md` — today at top.
- Honest status in `project-*.md` — what actually works, not the roadmap.

### 4. Save
- Create `context-stack/` if missing. Write the three context files into it.
- Write `CLAUDE.md` and `AGENTS.md` at the project root with the wrapper template.
- If any file already exists: don't overwrite silently. Show diff and ask, OR write as `<name>.new.md` for me to merge.

### 5. Debrief

```
## Context Stack built

**Created:**
- context-stack/identity.md         (N lines, from: [sources])
- context-stack/project-<slug>.md   (N lines, from: [sources])
- context-stack/state.md            (N lines, from: [sources])
- CLAUDE.md / AGENTS.md             (auto-loader wrappers)

**Files read:** [paths]
**Files skipped:** (one-line reason each)
**[TBD] markers:** (what would resolve each)

**Daily load:** open and ask "Where did I leave off?" — wrappers auto-load the three files.
**Weekly reset:** ~10 min updating each file + archiving last week's `state.md`.
```

## Templates (fill placeholders)

### `identity.md`
```markdown
# Identity — [your name]
**Role:** [your role]
**Focus:** [what you're optimizing for now]

## Background
[One short paragraph. Operating experience that filters your decisions. Under 60 words.]

## Tools I use
- [AI tools, platforms, core software]

## How I want you to answer
- [Direct/detailed/sparse · prose/bullets · push back or default-agree · cite paths when relevant]

## Non-negotiables
- [Short imperatives. What should an AI never do?]
```

### `project-<slug>.md`
```markdown
# Project — [project name]
**Tagline:** [one sentence]
**Positioning:** [who · what problem]

## Why it exists
[2–3 sentences. Skip the marketing — write the honest version.]

## Status (as of [YYYY-MM-DD])
- **What's live:** [what actually works]
- **What's in progress:** [partial · honest completion estimate]
- **What's next phase:** [coming up]

## Key decisions made
- [Decisions that recur in conversations.]

## Stakeholders
- [Name · role · how they engage]

## Deadlines / milestones
- [Date · what's due]

## Related artifacts (linked, not copied)
- `[path or URL]` — [what it is]
```

### `state.md`
```markdown
# State — week of [YYYY-MM-DD]

## This week
### [Day · YYYY-MM-DD] — [one-line theme]
- [what you shipped, touched, decided]

## Blocked
- **[short name]** — [what's blocking · what would unblock · who you're waiting on]

## Next (priority order)
1. [specific next move]
2. [next]
3. [next]

## Open threads (not blocking, not scheduled)
- [things on your mind that aren't active yet]

## Decisions made this week
- [decisions worth remembering past Monday]
```

### `CLAUDE.md` and `AGENTS.md` (Path A wrapper — same content in both)
```markdown
# Project context — auto-loaded

At session start, read and apply these files as authoritative working memory:

- `context-stack/identity.md` — how I want you to answer
- `context-stack/project-<slug>.md` — brief, status, decisions, stakeholders
- `context-stack/state.md` — this week's working memory

Follow the guidance in each. When I update any of these mid-session, re-read them.

If `state.md` was last updated more than 7 days ago, flag it before answering — it may be stale.
```

## Principles
- Safety first. Every read, every write logged. Nothing outside scope.
- Scaffold, don't solve. Build a starting point. I'll refine.
- The debrief is the product — tell me what you couldn't do.
- Weekly reset is my job, not yours.

Start with Step 1 (silent inventory), then Step 2 (your questions).
B · In a cloud Project
› Show the Path B prompt // markdown blocks · paste into your Project
context-stack-builder · path B.md 
Act as my Context Stack builder.

## Safety contract (Path B — cloud Project)

- Read only from what I've uploaded to this Project / attached to this chat.
- Output everything as labeled markdown code blocks. I'll save them to my Project myself.
- Skip credentials (keys, passwords, tokens) and anything marked confidential. If unsure, skip and note.
- Log every file you read in the debrief.

## What to build

Three files, output as labeled markdown code blocks (I'll paste them into my Project):

1. **`identity.md`** — who I am, how I want you to answer, my non-negotiables.
2. **`project-<slug>.md`** — this project's brief (goals, status, decisions, stakeholders, deadline). Lowercase slug, e.g. `project-feedrunner.md`.
3. **`state.md`** — this week's working memory.

(No wrapper files — those are a Path A filesystem thing. In a cloud Project, the assistant remembers files automatically once they're in the Project.)

## Steps

### 1. Inventory silently
Scan the uploaded files. Don't tell me what you found yet. Priority:
- High: charters, briefs, PRDs, scope docs, meeting notes, decisions.
- Medium: PDFs, DOCX, PPTX (data-extract), exported chats.
- Low: filenames as weak signal.

Skip secrets.

### 2. Ask 3–5 short questions
Batch in one message. Typical:
1. My role.
2. How I want you to answer (direct/detailed/sparse, prose/bullets, push back).
3. The one thing blocking or driving this week.
4. The last significant decision.
5. Who else is involved.

### 3. Draft the three files
Use the templates below. Rules:
- Specific beats complete.
- `[TBD: short reason]` when signal is weak. Don't invent.
- Recent-first in `state.md`.
- Honest status in `project-*.md` — what actually works.

### 4. Output as markdown blocks
Three labeled code blocks I can copy and save to my Project — one per file:

    ```markdown
    [file: identity.md]
    ... contents ...
    ```

    ```markdown
    [file: project-<slug>.md]
    ... contents ...
    ```

    ```markdown
    [file: state.md]
    ... contents ...
    ```

### 5. Debrief

```
## Context Stack built

**Output:** three files above, as markdown blocks.
**Files read:** [paths]
**Files skipped:** (one-line reason each)
**[TBD] markers:** (what would resolve each)

**Next:** save the three files to my Project (Claude Projects, ChatGPT Projects, Gemini, M365 Copilot, etc.). The Project will remember them across all future chats — I won't need to re-attach.

**To start a new session:** open the Project and ask "Where did I leave off?"
**Weekly reset:** ~10 min updating each file. Re-upload `state.md` to refresh.
```

## Templates (fill placeholders)

### `identity.md`
```markdown
# Identity — [your name]
**Role:** [your role]
**Focus:** [what you're optimizing for now]

## Background
[One short paragraph. Operating experience that filters your decisions. Under 60 words.]

## Tools I use
- [AI tools, platforms, core software]

## How I want you to answer
- [Direct/detailed/sparse · prose/bullets · push back or default-agree · cite paths when relevant]

## Non-negotiables
- [Short imperatives. What should an AI never do?]
```

### `project-<slug>.md`
```markdown
# Project — [project name]
**Tagline:** [one sentence]
**Positioning:** [who · what problem]

## Why it exists
[2–3 sentences. Skip the marketing — write the honest version.]

## Status (as of [YYYY-MM-DD])
- **What's live:** [what actually works]
- **What's in progress:** [partial · honest completion estimate]
- **What's next phase:** [coming up]

## Key decisions made
- [Decisions that recur in conversations.]

## Stakeholders
- [Name · role · how they engage]

## Deadlines / milestones
- [Date · what's due]

## Related artifacts (linked, not copied)
- `[path or URL]` — [what it is]
```

### `state.md`
```markdown
# State — week of [YYYY-MM-DD]

## This week
### [Day · YYYY-MM-DD] — [one-line theme]
- [what you shipped, touched, decided]

## Blocked
- **[short name]** — [what's blocking · what would unblock · who you're waiting on]

## Next (priority order)
1. [specific next move]
2. [next]
3. [next]

## Open threads (not blocking, not scheduled)
- [things on your mind that aren't active yet]

## Decisions made this week
- [decisions worth remembering past Monday]
```

## Principles
- Safety first.
- Scaffold, don't solve. Build a starting point. I'll refine.
- The debrief is the product — tell me what you couldn't do.
- Weekly reset is my job, not yours.

Start with Step 1 (silent inventory), then Step 2 (your questions).
C · In any chat
› Show the Path C prompt // minimal · save locally, re-attach next session
context-stack-builder · path C.md 
Act as my Context Stack builder.

## Safety contract (Path C — any chat)

- Read only from what I've attached to this chat.
- Output everything as labeled markdown code blocks. I'll save them locally and re-attach next session.
- Skip credentials (keys, passwords, tokens) and anything marked confidential. If unsure, skip and note.
- Log every file you read in the debrief.

## What to build

Three files, output as labeled markdown code blocks (I'll save them locally, re-attach next time):

1. **`identity.md`** — who I am, how I want you to answer, my non-negotiables.
2. **`project-<slug>.md`** — this project's brief (goals, status, decisions, stakeholders, deadline). Lowercase slug, e.g. `project-feedrunner.md`.
3. **`state.md`** — this week's working memory.

(No wrapper files, no Project upload. The three files ARE my portable Second Brain — drag them into the chat at session start.)

## Steps

### 1. Inventory silently
Scan the files I attached. Priority:
- High: charters, briefs, PRDs, meeting notes, decisions.
- Medium: PDFs/DOCX/PPTX (data-extract).
- Low: filenames as weak signal.

Skip secrets.

### 2. Ask 3–5 short questions
Batch in one message. Typical:
1. My role.
2. How I want you to answer (direct/detailed/sparse, push back or default-agree).
3. The one thing blocking or driving this week.
4. The last significant decision.
5. Who else is involved.

### 3. Draft the three files
- Specific beats complete.
- `[TBD: short reason]` when signal is weak. Don't invent.
- Recent-first in `state.md`.
- Honest status — what actually works.

### 4. Output as markdown blocks
Three labeled code blocks I can copy and save locally — one per file:

    ```markdown
    [file: identity.md]
    ... contents ...
    ```

    ```markdown
    [file: project-<slug>.md]
    ... contents ...
    ```

    ```markdown
    [file: state.md]
    ... contents ...
    ```

### 5. Debrief

```
## Context Stack built

**Output:** three files above. Save them locally as `identity.md`, `project-<slug>.md`, `state.md`.
**Files read:** [paths]
**Files skipped:** (reason each)
**[TBD] markers:** (what would resolve each)

**Next session:** drag the three files into a new chat, then ask "Load these. Where did I leave off?" That's it — no Project, no install.
**Weekly reset:** ~10 min updating each file. Save the new `state.md`, archive the old one.
```

## Templates (fill placeholders)

### `identity.md`
```markdown
# Identity — [your name]
**Role:** [your role]
**Focus:** [what you're optimizing for now]

## Background
[One short paragraph. Operating experience that filters your decisions. Under 60 words.]

## Tools I use
- [AI tools, platforms, core software]

## How I want you to answer
- [Direct/detailed/sparse · prose/bullets · push back or default-agree]

## Non-negotiables
- [Short imperatives. What should an AI never do?]
```

### `project-<slug>.md`
```markdown
# Project — [project name]
**Tagline:** [one sentence]
**Positioning:** [who · what problem]

## Why it exists
[2–3 sentences. Skip the marketing — write the honest version.]

## Status (as of [YYYY-MM-DD])
- **What's live:** [what actually works]
- **What's in progress:** [partial · honest completion estimate]
- **What's next phase:** [coming up]

## Key decisions made
- [Decisions that recur in conversations.]

## Stakeholders
- [Name · role · how they engage]

## Deadlines / milestones
- [Date · what's due]

## Related artifacts (linked, not copied)
- `[path or URL]` — [what it is]
```

### `state.md`
```markdown
# State — week of [YYYY-MM-DD]

## This week
### [Day · YYYY-MM-DD] — [one-line theme]
- [what you shipped, touched, decided]

## Blocked
- **[short name]** — [what's blocking · what would unblock · who you're waiting on]

## Next (priority order)
1. [specific next move]
2. [next]
3. [next]

## Open threads (not blocking, not scheduled)
- [things on your mind that aren't active yet]

## Decisions made this week
- [decisions worth remembering past Monday]
```

## Principles
- Safety first.
- Scaffold, don't solve.
- The debrief is the product — tell me what you couldn't do.
- Weekly reset is my job, not yours.

Start with Step 1 (silent inventory), then Step 2 (your questions).
// On GitHub Copilot · auto-attach your Second Brain in VS Code

If your work happens in VS Code, GitHub Copilot can auto-load your Second Brain on every chat — no copy/paste, no Project setup. The trick: .github/copilot-instructions.md is attached to every Copilot Chat in the repo, by default. Point it at your identity.md + state.md and Copilot opens already knowing who you are.

// github copilot · vs code

Your context-stack as .github/copilot-instructions.md — Path A semantics, IDE-native.

Copilot Chat reads .github/copilot-instructions.md at the start of every chat in the repo. If your context-stack lives in a private "second-brain" repo (or alongside your code), that file becomes your auto-attached identity + state — no manual context loading per session.

// step 1 · enable

Turn on instructions

VS Code › settings › search github.copilot.chat.codeGeneration.useInstructionFiles › enable. (On by default in newer Copilot Chat builds.)

// step 2 · reference

Point at your stack

Create .github/copilot-instructions.md at your repo root:

# Working with me

You are pair-programming with [your name].

## Identity
[paste identity.md]

## Current state
See ./state.md for what I'm working
on this week.

## Style
[paste style.md or wrappers]
// step 3 · verify

Confirm auto-load

Open Copilot Chat. Ask "What do you know about me?" Copilot should answer with content from identity.md. If not, check the chat header — instructions files appear under "References" when active.

Why this matters: the same context that powers your Path A wrappers (CLAUDE.md / AGENTS.md) now powers GitHub Copilot. One source, three activation surfaces: Claude Code reads CLAUDE.md, agentic CLIs read AGENTS.md, GitHub Copilot reads .github/copilot-instructions.md. For private second-brains, keep the repo private; the instructions file never leaves the repo's read scope.

// Keep it fresh

Your Second Brain is only as useful as your state.md. Two habits keep it alive:

// daily · ~60 sec

Open and ask "where did I leave off?"

Path A wrappers auto-load the three files. Path B/C: re-attach state.md at session start. Your AI catches you up on what's blocked and what's next.

// monday · ~10 min

Reset

Update each project-*.md with last week's decisions. Archive last week's state.md. Start a fresh one. The bundle PDF has the click-by-click drill.