Module 3 · Cowork Setup + Agent Build Kit

Build Your AI Agents

This module runs on Claude Cowork, the agent workspace in the Claude desktop app. Unlike a normal chat, it reads a real folder on your computer, follows your instructions on every task, and saves real files. Part A sets it up properly. Part B gives you three life agents.

Why an agent workspace, not just a chat? Cowork reads a real folder on your computer, follows your global instructions before every task, and saves deliverables to disk. That's what makes these three feel like assistants instead of chats you re-explain yourself to.

Part A: Set up Claude Cowork

Work through these four steps once. ~15 minutes.

1

Get access

2

Build your folder structure

This is the home your agents live in. Make these folders and files inside Workflow Vault Cowork:

Workflow Vault Cowork/ ├── ABOUT ME/ │ ├── about-me.md │ ├── writing-style.md │ └── my-life.md ├── OUTPUTS/ ← agents save deliverables here └── TEMPLATES/ ← reusable prompts & templates
ABOUT ME/about-me.md Who you are, how you think, how you want Claude to work with you. Keep under ~2,000 tokens.
Build it by interview, paste this into Cowork:
Interview me to build my about-me.md file for Claude Cowork. Ask one question at a time about: who I am, the roles I hold, how my brain works best, how I want you to work with me, and what "good" output looks like to me. Keep the result under 2,000 tokens. When we're done, output the finished about-me.md in a single code block I can copy.
ABOUT ME/writing-style.md The words and patterns you hate, AI clichés to ban, formatting rules. A list of hard rules.
ABOUT ME/my-life.md Your roles, goals, constraints, and what you're saying no to. Adapted from your Module 1 BUILD SPEC.
Or build it by interview, paste this into Cowork:
Interview me to build my my-life.md file. Ask one question at a time about my roles, what I'm working toward in the next 1–3 months, my hard constraints (time, energy, money, childcare), my non-negotiables, and what I'm deliberately saying no to right now. When done, output the finished my-life.md in a single code block.
3

Set your global instructions

Paste this into Settings → Cowork → Edit Global Instructions. It tells every agent to read your ABOUT ME files first and where to save things.

Before every task, read every file in ABOUT ME/:
- about-me.md: who I am, how I think, what I need
- writing-style.md: how I want you to write, what to avoid
- my-life.md: my roles, my constraints, what I'm working toward

Never read OUTPUTS/ or TEMPLATES/ unless I point you to a specific file.
Save all deliverables in OUTPUTS/ under a subfolder named after the task or project.
If the brief is unclear, ask me questions before starting. Don't fill gaps with filler.
4

Token-saving rules

Cowork reads your files on every task, so keep things lean. Check these off as habits:

Full cheat sheet lives in Cowork Quick Reference.

5

Connect your tools (optional, but this is where it gets powerful)

You can fill every field in this product by hand. Or connect the tools you already live in, and your agents read the real thing live. No copy-paste, always current.

📅 Google Calendar

Your Daily Planner agent sees when you're actually booked, finds your open focus windows, and stops scheduling over real commitments. Instead of painting busy times by hand in Module 2, ask: "Look at my calendar this week and tell me where my free deep-work blocks are."

📔 Notion

Your agents read your projects, tasks, and notes. Instead of typing out what each role or project is, ask: "Pull my active projects from Notion and tell me what each one needs this week."

How to connect (about 2 minutes)

  1. In the Claude desktop app, open Settings, then Connectors.
  2. Click the + in the Connectors header, search for Google Calendar and Notion, and connect each (about 90 seconds apiece). You sign in and approve access once.
  3. Back in a Cowork session, just ask in plain language: "Check my Google Calendar for next week" or "Pull my projects from Notion." The agent uses your live data.

Anywhere this product asks you to fill something in, that's also a place you can connect a tool instead. The two work together: connect what you can, type the rest.

Part B: Your three life agents

Each agent: copy the system prompt, generate + download its context files, deploy in Cowork. Hit Generate to personalize the files from your Module 1 + 2 answers.

1
Agent 1

Daily Planner Agent

Turns your roles, schedule, and non-negotiables into a realistic plan, and gives you one next action at a time instead of a list.

System prompt
You are my Daily Planner Agent in Claude Cowork. Before every task, read ABOUT ME/ and these files: DAILY-RULES.md, SCHEDULE.md, PRIORITIES.md.

How to work with me:
1. Ask for today's brain dump (or read what I paste).
2. Map tasks onto SCHEDULE.md, hardest task in my peak window.
3. Never schedule over anything in DAILY-RULES.md.
4. Give me ONE next action at a time.
5. If I'm overcommitted, say so and tell me what to drop.

Save any plan in OUTPUTS/daily-plan/. Tone: direct, warm, no filler.
(Hit "Generate My Agent Files" to drop your roles in here.)
Context files (generate, then download into ABOUT ME/ or alongside this agent)
DAILY-RULES.mdNon-negotiables + energy map
SCHEDULE.mdYour weekly time blocks
PRIORITIES.mdHow you triage when everything's urgent
How to deploy in Cowork
Steps
  1. In Cowork, save the three files into your Workflow Vault Cowork folder (a Daily Planner/ subfolder is tidy).
  2. Start a session and paste the system prompt as your first message.
  3. Say: “Read my files, then ask for today's brain dump.”
2
Agent 2

Deep Work / Focus Agent

For any cognitively demanding work that needs protecting. Tell it your work type and it adjusts.

System prompt
You are my Deep Work / Focus Agent in Claude Cowork. Before every session read ABOUT ME/ plus DEEP-WORK-RULES.md and WORK-QUALITY.md.

At the start: help me pick ONE focus target and a time box. During: if I drift, name it; protect the block. Watch for my fatigue signals and tell me to stop when they show. After: prompt me to fill SESSION-LOG.md.

Tone: crisp and protective, sustained quality over raw speed.
(Pick your work type above, then Generate to tailor this.)
Context files
DEEP-WORK-RULES.mdHow you protect + enter focus
WORK-QUALITY.mdWhat good output looks like for your work
SESSION-LOG.mdDaily deep-work tracker
How to deploy in Cowork
Steps
  1. Save the three files into a Deep Work/ subfolder.
  2. Paste the system prompt to start a focus session.
  3. Say: “Help me pick one target and a time box, then protect it.”
3
Agent 3

Life Admin Agent

Household, family, and the logistics that run in the background, appointments, school calendar, pickup windows.

Personal logistics, errands, appointments, recurring personal tasks, and the background life-admin that eats your attention.

System prompt
You are my Life Admin Agent in Claude Cowork. Before every task read ABOUT ME/ plus LIFE-RULES.md, CALENDAR-ANCHORS.md, and QUICK-WINS.md.

Each check-in: surface what's coming in the next 48 hours that needs action, especially anything that collides with a work block. Turn my vague worries into concrete next actions. When I have a free gap, hand me ONE quick win, never the whole list. Keep CALENDAR-ANCHORS.md current.

Tone: calm, practical, one or two next actions max.
(Hit Generate to tailor the files to your situation.)
Context files
LIFE-RULES.mdRecurring tasks, contacts, household rules
CALENDAR-ANCHORS.mdFixed events that cannot move
QUICK-WINS.mdSub-10-min tasks to fill gaps
How to deploy in Cowork
Steps
  1. Save the three files into a Life Admin/ subfolder.
  2. Paste the system prompt to start.
  3. Say: “Give me the next 48-hour heads-up.”

Want every blank template at once? Grab them all from the Cowork Quick Reference as a single ZIP, or from the /templates/ folder of this download.