From Zero to Claude Code Pro in 90 Minutes (For PMs)
From new install to PM Copilot in 90 minutes
Follow along with the live session. By the end, you'll have synthesized research, built a prototype, and set up your own PM operating system.
Download Starter FilesBefore We Start
You need:
Cursor (or VS Code)
Any code editor works – we just need to see the files Claude creates
Download Cursor →Starter Files
Unzip to a folder you can find easily. This is your "messy inbox" from your new job.
Download starter files →The Scenario
You just joined Faculty Products as a PM. The company makes GradeFlow, a course management platform used by 200+ universities.
The PM role has been empty for months. Your boss hands you a folder full of research files, company docs, and notes. "Figure out what's important and ship something by end of week."
We're going to do it in 90 minutes.
Your First Day
Getting up to speed and shipping something
Read & Summarize the Company Overview
Long company doc. Need to get up to speed fast.
Summarize @company-overview.md - what do I need to know as a new PM?
What you learned
- •`@filename` references files directly - no copy-pasting
- •Claude works WITH your files, not in a separate chat window
- •You can ask follow-ups without re-uploading anything
📖 GradeFlow serves 200+ universities. TAs are supposed to be 'our differentiator' but adoption is 'not great.' One champion professor said 'GradeFlow doesn't solve my real problem.'
Synthesize the User Research
12 interview transcripts from professors and TAs. Never synthesized.
Read all the files in @research and give me the top 3 pain points across these interviews. Save the synthesis to user-research-synthesis.md
What you learned
- •Claude reads across multiple files at once
- •Creates real files that persist in your project
- •Pattern recognition across documents in seconds
📖 12 interviews, one pattern: TAs are drowning. James fell 60 hours behind – no one noticed until students complained. Emily and Marcus graded the same essay: B+ and C+. Everyone's grading in isolation.
Create Your Learning Summary
Write up what you learned for your manager.
Based on @q1-goals.md and @user-research-synthesis.md, write a summary of what I've learned. Save it as learning-summary.md
What you learned
- •Combine multiple sources into one output
- •Work accumulates as real files, not chat history
📖 Three themes emerged: workload visibility, grading consistency, TA isolation. The research never mentioned AI – they wanted human problems solved first.
Format It for the Boss
Manager wants a formal 1-pager using the company template.
Put @learning-summary.md into the format from @one-pager.md. Save it as one-pager-learning-summary.md
What you learned
- •Template-based transformation: one file = structure, another = content
- •Your templates encode company standards
- •Separate WHAT you're saying from HOW it's formatted
Make It Sound Human
Reads like AI wrote it. Need to clean it up.
Rewrite @one-pager-learning-summary.md using the style guidelines in @writing-style.md
What you learned
- •Apply any set of rules to any content
- •Build a library of transformation prompts
- •Scale your writing without it becoming generic
📖 First draft was functional but generic. After applying the style guide, it matched Faculty Products' voice – empathetic, practical, professor-first.
Share It Three Ways
Different stakeholders want different formats.
Take @one-pager-learning-summary.md and create three versions: 1. Use @slack-update.md format 2. Use @email-summary.md format 3. Use @company-wiki.md format
What you learned
- •One source of truth, many distribution formats
- •Write once, share everywhere
- •Templates for every channel you use
The Full Workflow
Everything we just did could be one command.
Open workflows/onboarding-to-update.md in your editor – it chains all the previous steps together:
- Synthesize research
- Create summary
- Format as 1-pager
- Humanize
- Distribute to all channels
This is the pattern: chain steps together in a markdown file, then run it whenever you need it.
What you learned
- •Workflows = multi-step processes defined in markdown
- •Each piece can be modified independently (different templates, different sources)
- •This is the core pattern: combine and transform information in repeatable ways
📖 One file chains the whole process: research → summary → format → humanize → distribute. Write once, run anytime.
Create Requirements
Grading inconsistency is the big problem. Let's spec a solution.
Based on @user-research-synthesis.md and @q1-goals.md, create a 1-page requirements doc for a TA Workload Tracker. Use the AskUserQuestions tool to ask me lots of questions about it. Save as requirements.md
What you learned
- •Multiple context files inform a single output
- •Claude asks YOU questions to fill gaps
- •Go from research pile to actionable doc in minutes
📖 Q1 goals include 'Ship AI Feedback Assistant' with 50% adoption target. The research never mentioned AI once. There's a gap between the roadmap and what users actually need.
Get Feedback from Three Perspectives
Pressure-test before the real review meeting.
Review @requirements.md from three perspectives using @skeptical-engineer.md, @busy-executive.md, and @ux-designer.md. Save as @req-review-ta-dash.md.
What you learned
- •Simulate stakeholder perspectives before the actual meeting
- •No waiting for calendars
- •Walk into meetings prepared
📖 The Skeptical Engineer asked about backend load. The Busy Executive wanted timeline to value. Questions you hadn't thought to ask yourself.
Finalize Requirements
Incorporate feedback and lock it down.
Use your AskUserQuestions to help me walk through the feedback. Then update @requirements.md
What you learned
- •You make the decisions, Claude executes
- •Interactive refinement keeps you in control
- •Your judgment shapes the final output
- •Leave comments in docs with {{like this}} so Claude can address them all at once
- •Use /ide to connect to your editor so Claude can see what you see
Build the Prototype
We have requirements. Let's build.
Switch to Plan Mode first.
Based on @brandguidelines.md and @requirements.md, build a TA Workload Tracker prototype. Use HTML/CSS/JS so I can open it in a browser.
Watch for Plan Mode
- • Claude shows its approach before building
- • You approve, adjust, or redirect
- • Course-correct before any code is written
What you learned
- •Plan mode = see the approach before execution
- •You're not hoping it does the right thing
- •Mirrors working with engineers: "Here's my plan" → "Looks good"
📖 Blank screen to working prototype in 10 minutes. The prototype pulls from brand guidelines, matches requirements, and is ready to demo.
While It Builds: Parallel Sessions
Don't wait around. Start another task.
- • Open a new terminal tab
- • Run
claude - • Start a different task while the prototype builds
What you learned
- •Multiple Claude sessions run simultaneously
- •Never blocked waiting
- •Like having multiple team members
Connect Your Real Tools (MCPs)
How do I get my ACTUAL work files in here? Not just local files, but Google Drive, Jira, Notion...
The answer: MCPs (Model Context Protocol) – plugins that connect Claude to external tools.
Live demo: Linear
claude mcp add --transport http linear-server https://mcp.linear.app/mcpCommon MCPs for PMs:
| Tool | What it does | Setup Link |
|---|---|---|
| Google Workspace | Drive, Docs, Gmail, Calendar, Sheets | GitHub |
| Jira + Confluence | Official Atlassian MCP | GitHub |
| Notion | Pages, databases, wikis | GitHub |
| Linear | Issues, projects, roadmaps | Docs |
| Slack | Messages, channels, search | GitHub |
What you learned
- •MCPs = plugins that connect Claude to external services
- •Your real work lives in these tools - now Claude can access it
- •Check with your company's security policies before enabling
Build Your Operating System
Making Claude YOUR copilot
Slash Commands: See One
You've been typing patterns repeatedly. What if you could save them?
- • Type
/in Claude to see available commands - • Open
.claude/commands/humanize.mdin your editor - • It's just a markdown file with instructions
/humanize @stakeholder-update-draft.md
What you learned
- •Slash commands = saved prompts as reusable tools
- •No more typing the same patterns
- •Build a personal toolkit
📖 The /humanize command is just a markdown file. 30 seconds to create, hours saved over time.
Slash Commands: Build One
Save that multi-perspective review as a command.
Test it:
/prd-review @requirements.md
What you learned
- •Create commands in seconds, no code required
- •Any workflow becomes a one-word trigger
- •Every pain point becomes a solved problem
CLAUDE.md: Project Memory
Claude doesn't remember your context between sessions. Let's fix that.
Step 1: Ask without context
Tell me about this company and what we should prioritize
Notice: Claude gives generic advice – it doesn't know anything about Faculty Products.
Step 2: Create project memory
Clear the session with /clear, then:
Create a CLAUDE.md file that describes this workspace: we're PMs at Faculty Products working on GradeFlow. Include the key user insights from @user-research-synthesis.md and @company-overview.md
Step 3: Ask the same question
Clear again with /clear, then ask the same question:
Tell me about this company and what we should prioritize
Now Claude answers with context – it knows the company, the users, the research findings.
What you learned
- •CLAUDE.md = persistent memory loaded automatically on every session
- •No more "let me give you context..." every time
- •Your copilot has institutional memory
📖 Same question, different answers. Before: generic AI advice. After: recommendations grounded in GradeFlow's actual users and constraints.
The Mental Model
Your PM Workspace Structure:
pm-copilot-os/
├── CLAUDE.md # What Claude should KNOW
├── README.md
├── tasks.md
├── backlog.md
├── .claude/
│ └── commands/ # What Claude should DO
├── context/
├── meetings/
│ ├── standups/
│ ├── 1-on-1s/
│ ├── planning/
│ └── reviews/
├── projects/
├── templates/
│ └── personas/
├── tools/
├── workflows/
└── _temp/Key concepts:
- One folder = one workspace –
cdinto your folder, thenclaude - CLAUDE.md = what Claude knows – role, context, terminology
- Slash commands = what Claude does – repeatable workflows
- Multiple workspaces – PM work, side projects, personal stuff
Get the templates:
| Template | What's inside | Download |
|---|---|---|
| PM Copilot OS (Example) | Full setup with Faculty Products content – CLAUDE.md, 11 slash commands, templates, workflows, example project | Download |
| PM Copilot OS (Blank) | Clean starter – just the folder structure, ready for your content | Download |
📖 Claude isn't a chatbot to talk to. It's an operating system that knows your job, your company, your workflows.
What You Built
In 90 minutes, you:
Your boss gave you one week. You did it in 90 minutes.
Next Steps
Resources
- Google Workspace MCP Setup – Connect Google Drive, Docs, etc.
- Notion MCP Setup – Connect your Notion workspace
- Full Stack PM Newsletter – Weekly tips and workflows
Quick Reference
Essential Commands
| What | Command |
|---|---|
| Reference a file | @filename.md |
| Reference a folder | @foldername |
| Run a slash command | /commandname |
| Switch models | /model |
| Clear conversation | /clear |
| Undo last action | Escape |
| Rewind further | Escape × 2 |
File Types You'll Use
| File | Purpose |
|---|---|
CLAUDE.md | Project context (auto-loaded) |
.claude/commands/*.md | Slash command definitions |
templates/*.md | Output format templates |
templates/personas/*.md | Stakeholder perspectives |
The Pattern
- Reference context files with
@ - Transform using templates
- Review from multiple perspectives
- Save workflows as slash commands
- Build institutional memory in CLAUDE.md
FAQ
Do I need to know how to code?
No. Everything we do uses natural language. Claude writes the code.
Can I use VS Code instead of Cursor?
Yes. Any editor works – we just need to see the files Claude creates.
What if I fall behind during the session?
Recording will be available. This page stays up as reference.
How do I connect my real work tools?
MCP integrations (Google Drive, Notion, Jira, etc.) – links in Resources section.
Can I share my slash commands with my team?
Yes! They're just markdown files. Commit them to your repo or share directly.