claudecodeguide.dev

Stop drowning in
meeting prep.

I built this because I was spending 60% of my day on status updates instead of actual product work. Works with ChatGPT, Claude, Gemini. Gets a lot more powerful with Claude Code.

New guides, when they ship

One email, roughly weekly. CLAUDE.md templates, workflows I actually use, and the cut-for-length stuff that does not make the public guides. One-click unsubscribe.

Or follow on Substack

Live example

pm-pilot : meeting prep

Works with whatever AI you're already using

ChatGPT
Claude
Gemini
Cursor
VS Code
01

What actually changes

I didn't want a tool that thinks for me. I wanted one that handles the busywork so I could actually think.

Meeting prep

45 min digging through Jira, Slack, and last week's notes. Then the meeting starts.

30 seconds. A brief drops in your terminal. You walk in actually ready.

Weekly status

2 hours chasing people for updates and guessing at sprint progress

Pulled from real Jira data, formatted, ready to send. Takes a minute.

PRD writing

Staring at a blank Confluence template for 40 minutes, writing nothing

Braindump first, structure second. You think out loud; it organises the output.

Market sizing

Days in spreadsheets with numbers you half-trust

TAM, SAM, SOM with explicit assumptions baked in. Takes minutes.

03

Five skills worth trying first

Each one is a markdown file. Read it, edit it, fork it. Nothing is hidden.

01
/meeting-prep

Checks Jira, Slack, and your calendar, then gives you a brief. Who owes what, what got missed, what to ask. I use this before every 1:1.

02
/prd

Asks you to brain-dump first. No template until the thinking is done. Genuinely prevents the thing where you write a whole PRD that says nothing.

03
/weekly-status

Hits Jira, counts what closed, flags the blockers, writes the update. You just review and send.

04
/market-sizing

TAM, SAM, SOM with all the assumptions written out. Saves you from the spreadsheet that nobody believes anyway.

05
/people-sync

Reads transcripts from Granola and updates your stakeholder notes automatically. No more forgetting what you promised last week.

/weekly-status, running live

pm-pilot : weekly status
02

Start where you are

Never opened a terminal? Fine. You can get value from PM Pilot today, without installing anything. Upgrade whenever you're ready.

01

Any AI chat, right now

Zero install

Grab any skill file from GitHub and paste it into ChatGPT, Claude, or Gemini. No terminal. No account. You can try this in the next 2 minutes.

02

Claude Desktop

5 min setup

Create a project in Claude Desktop and paste any skill into the project instructions. No terminal needed. Skills stick across every conversation.

03

Claude Code CLI

Live Jira, Slack, Calendar

This is where it gets serious. PM Pilot reads live data: real tickets, real Slack threads, real deadlines. I use this every day.

05

Three ways in

Pick what fits where you are right now. You can always go deeper later.

01

Zero install

ChatGPT, Gemini, claude.ai

Open claude.ai, ChatGPT, or Gemini. Copy any skill file and paste it straight into the chat. That's it.

No Jira or Slack yet, but you'll immediately see how structured the output is. Worth 2 minutes to try.

Open the meeting-prep skill file
02

Claude Desktop

5 minute setup

  1. Download the Claude Desktop app
  2. Open it. Go to Settings → Projects
  3. Create a project. Call it "PM Pilot" or whatever you like.
  4. Paste any skill file into the project instructions

Done. Skills persist between sessions. No terminal, ever.

03

Claude Code CLI

Full power

This is the real thing. Live Jira, Slack threads, calendar, meeting transcripts. I built PM Pilot for this setup; it's what I actually use.

Follow the Claude Code setup guide to get started, then run:

git clone https://github.com/mshadmanrahman/pm-pilot.git
cd pm-pilot
mkdir -p ~/.claude/skills ~/.claude/rules ~/.claude/agents ~/.claude/commands ~/.claude/memory
cp -r skills/* ~/.claude/skills/
cp -r rules/* ~/.claude/rules/
cp -r agents/* ~/.claude/agents/
cp -r commands/* ~/.claude/commands/
cp memory/MEMORY-TEMPLATE.md ~/.claude/memory/MEMORY.md

Then run /configure-pm-pilot inside Claude Code and you're set.

No git installed? Get it here first.

06

Connect your work tools

Claude Code CLI only. This is where PM Pilot goes from "useful" to "I can't work without it."

Jira / Confluence

meeting-prep, weekly-status, deep-context

Atlassian MCP

Slack

meeting-prep, weekly-status, deep-context

Slack MCP

Google Calendar

meeting-prep (knows your schedule)

Google Calendar MCP

GitHub

weekly-status, code context

GitHub MCP

Granola

people-sync (meeting transcripts)

Get Granola

No integrations yet? Skills like market-sizing, prd, prioritize, and critique don't need any of this. They work the moment you install PM Pilot.

07

It remembers things, so you don't have to repeat them

Every other AI tool starts fresh each session. PM Pilot builds a persistent knowledge base. You explain your context once. That's it.

How the memory is structured

~/.claude/memory/
  MEMORY.md              - always loaded, under 200 lines
  project_*.md           - one file per project you own
  feedback_*.md          - corrections (so Claude doesn't repeat mistakes)
  user_*.md              - your preferences and working style
  people/                - one file per key person
    sarah-chen.md
    marco-vidal.md
1

Day 1

Memory is empty. You explain your projects, your team, your preferences. Just this once.

5

Day 5

Projects, key people, and preferences are saved. You spend a lot less time re-explaining things.

15

Day 15

Meeting prep starts pulling stakeholder context automatically. Patterns from past sessions are baked in.

30

Day 30

Sessions start with full context loaded. You just say what you need. No backstory.

Here's the thing:every time you correct something, that correction gets saved. You say it once. It doesn't come up again.

08

Why this exists

I tracked my time for a week. More than half of it was status updates, meeting prep, and chasing context across Jira and Slack. That's not product work. That's admin with a fancier title.

So I built a system to eat the admin.

14 years of PM experience across startups and enterprise, compressed into reusable skills. The kind of setup I always wanted but never had time to build properly. Until I did.

01

Braindump before structure. Templates create false clarity. Thinking comes first; the skeleton follows.

02

Memory beats transcripts. A 200-line memory file is more useful than scrolling back through a 50,000-token conversation.

03

Skills load on demand. There are 25 of them, but only the ones you actually use ever touch your context window.

04

Rules stay compressed. Around 950 tokens total. Not 9,500.

05

Every session builds on the last. Corrections become saved rules. Meetings become stakeholder notes.