AI PM Workspace

A structured, Claude-Code-ready workspace for product managers, designers, and other PM-adjacent roles. Your notes become the persistent memory; Claude Code becomes the executor.

This is a template. Clone it, run /personalize --quick (three minutes), and you have a working AI workspace.

v2 is live (see CHANGELOG.md). Tasks now live in-repo at Dashboard/tasks.md - no external task manager required. The old Todoist-coupled v1 is available on the v1 tag if you need it.

Background reading:

• Your AI Has No Memory. Mine Does. - the architecture that inspired this template.
• The Four Claude Code Building Blocks - CLAUDE.md, MCP, skills, agents.
• Claude Has 5 Memory Layers - how Claude Code actually remembers you.

Get started in 5 minutes

Three steps. Two of them are skills.

1. Clone the template. Pick the path that matches how you work:

• No terminal: click the green "Use this template" button above, then clone with GitHub Desktop.
• Minimal terminal: open Claude Code in any empty folder and ask it: "Clone the template at https://github.com/aleksander-dytko/ai-pm-workspace into a new repo called my-pm-workspace on my GitHub account, then clone it locally."
• Full terminal: gh repo create my-pm-workspace --private --template aleksander-dytko/ai-pm-workspace && gh repo clone my-pm-workspace && cd my-pm-workspace

Full instructions with screenshots and gotchas: Quick start below and setup/SETUP-GUIDE.md.

2. Run /personalize --quick (3 minutes).

Open Claude Code in the cloned folder and type /personalize --quick. It asks three multiple-choice questions and one free-text question ("your name, your company in one sentence, your focus area"). Fills your identity into CLAUDE.md, trims the workspace to your role and audiences. You're ready to work.

3. Run /guide (eight modules, one per sitting).

/guide is an interactive learning path that walks you through every shipped skill using sample data that's already in the repo. No scrambling for a real meeting or a real decision to practice on. Role-adaptive so it reframes examples for designers, analysts, and engineers too.

After that, try the skills on real work. When you want more context on board, run /personalize --deep (10 min) to add MCP detection, active initiatives, and people profiles.

What you get

• 12 ready-to-use skills covering daily PM work, craft artifacts, onboarding, and feedback (see the full list below).
• A structured vault: journals, meeting notes, decisions, dashboards, initiatives, epics, research.
• CLAUDE.md pre-configured with quality standards, interaction rules, and a skill dispatch table.
• In-repo task management via Dashboard/tasks.md - no Todoist or external tool required.
• An interactive /guide that walks new users through all the shipped skills in eight modules, using sample data bundled with the template.
• Templates + personas + samples ready to use: epic template, initiative template, three generic personas, six sample data files.

How it fits together

+-------------------+    +---------------------+    +-------------------+
|                   |    |                     |    |                   |
|   YOUR VAULT      |<-->|    CLAUDE CODE      |<-->|   MCP BRIDGES     |
|   (persistent     |    |    (executor)       |    |   (optional)      |
|    memory)        |    |                     |    |                   |
|                   |    |   Skill categories: |    |   - GitHub        |
|  - Daily journal  |    |    * Core daily     |    |   - Your docs     |
|  - Meeting notes  |    |    * Craft          |    |   - [Your tools]  |
|  - Decisions      |    |    * Onboarding     |    |                   |
|  - Epics          |    |    * Feedback       |    |                   |
|  - Research       |    |                     |    |                   |
|  - Dashboard/     |    |                     |    |                   |
|    tasks.md       |    |                     |    |                   |
+-------------------+    +---------------------+    +-------------------+
        ^                         |
        |       writes back       |
        +-------------------------+

• Your vault accumulates knowledge in Markdown. Everything searchable, versionable, yours.
• Claude Code reads your vault and runs skills. Skills are repeatable workflows encoded as Markdown files.
• MCP bridges (optional) connect Claude Code to external tools like GitHub.

The skills

Core daily (5)

Craft (4)

Onboarding (2)

Feedback (1)

Quick start

Prerequisites

• A Claude Pro ($20/mo), Max, or Teams subscription. The free plan does not include Claude Code.
• Claude Code installed. See setup/SETUP-GUIDE.md for three install paths.
• A GitHub account.
• (Optional) Obsidian if you want the vault UX.

Clone the template (three paths)

Pick the path that matches how you work.

Path 1: GitHub UI + GitHub Desktop (no terminal required)

Best for: designers, PMs, anyone who doesn't live in a terminal.

1. Click the green "Use this template" button at the top right of this repo page.
2. Choose "Create a new repository". Name it (e.g., my-pm-workspace), make it private, create.
3. Open GitHub Desktop. File > Clone repository > [your new repo].
4. Pick a local folder, click Clone.
5. Open Claude Code in that folder. Run /personalize. Done.

If you get stuck on the GitHub side of this, Hannah Stulberg's Tool School: GitHub 101 is a great primer for non-developers, and setup/SETUP-GUIDE.md has platform-specific install commands.

Path 2: Ask Claude Code to do it (minimal terminal)

Best for: PMs who have Claude Code installed but aren't fluent with git.

Open Claude Code in any empty folder and paste:

Clone the template at https://github.com/aleksander-dytko/ai-pm-workspace
into a new repo called my-pm-workspace on my GitHub account,
then clone that new repo locally and open it.

Claude will use the GitHub CLI to create the repo and clone it. If you don't have gh installed, Claude will guide you through installing it.

Then run /personalize.

Path 3: Git CLI (classic)

Best for: developers.

gh repo create my-pm-workspace --private --template aleksander-dytko/ai-pm-workspace
gh repo clone my-pm-workspace
cd my-pm-workspace
claude   # run Claude Code

Then run /personalize.

What /personalize does

• Quick (default, 3 min): asks your name, role, focus area, and team. Fills CLAUDE.md identity. Ships working skills immediately.
• Deep (10 min): everything above, plus MCP detection, initiative setup, and people profiles. Run /personalize --deep when you want more context.

Going deeper

Once /personalize --quick is done, you have two paths.

Try the skills one by one:

1. /today - morning planning or evening close-out.
2. /meeting - process a raw meeting note into a structured one.
3. /decision - document a product decision with full context.
4. /communicate - draft a Slack, email, or async update.
5. /weekly-plan - Sunday/Monday planning with an overplanning challenge.

Or run the interactive guide:

/guide

Eight modules, role-adaptive, using sample data shipped with the template. Covers all shipped skills plus how to extend your workspace with external skills. Good to do over a week, one module per day.

When you want more, /personalize --deep adds MCP integration, initiatives, and people profiles.

How CLAUDE.md works

CLAUDE.md is the single most important file in this setup. Claude Code reads it at the start of every session, so it always knows:

• Your identity (role, company, focus area)
• How your vault is organized (folders, naming conventions)
• Quality standards for notes, decisions, and communications
• Interaction rules (never-write-tasks-without-confirming, no em-dashes, etc.)
• Which skills exist and when to use them

The more context you put into CLAUDE.md, the less you need to explain in each conversation. Deep dive: Claude Has 5 Memory Layers.

Extending the workspace

Build your own skills

Skills are Markdown files in .claude/skills/[name]/SKILL.md. Each has:

1. Frontmatter: name and description (how Claude discovers the skill).
2. Workflow: step-by-step instructions.
3. Output format: what the result should look like.

Ideas to build yourself:

• /retro - sprint or quarterly retrospective from journals and meeting notes.
• /standup - daily standup summary from journal + today's plan.
• /health - vault health audit (unlinked notes, stale tasks, orphaned decisions).

Module 8 of /guide walks you through how to write your own or adopt one from the internet.

Port skills from elsewhere

There's a growing open-source library of PM skills you can borrow.

• phuryn/pm-skills - 60+ MIT-licensed skills covering discovery, strategy, GTM, analytics, and more. One of these (opportunity-solution-tree) already ships with this template as a worked attribution example.

To add another phuryn skill to your workspace, copy its folder into .claude/skills/, keep the attribution header, restart Claude Code.

Optional: sync tasks to Todoist

This template keeps tasks in-repo (Dashboard/tasks.md) so you can start without any external tool. If you want Todoist sync later, see docs/todoist-sync.md.

Philosophy

Three principles:

1. Context over prompts: when AI has deep context, the prompt barely matters. A simple "What should we do about X?" becomes a rich interaction because Claude already knows what X is.
2. Skills over conversations: repeatable workflows shouldn't require re-explaining every time. Skills encode your process once and execute it consistently.
3. Your vault, your memory: everything accumulates in Markdown files you own. No vendor lock-in, no cloud dependency for your knowledge.

Migrating from v1

If you cloned v1 (Todoist-coupled) and want to upgrade: see the migration notes in CHANGELOG.md. Short version: tasks now live in Dashboard/tasks.md instead of Todoist, and four skills (/end-of-day, /process-inbox, /slack-thread, and parts of /personalize) were either merged or removed. Your notes and journals are unchanged.

Contributing

This is an open template. If you build skills, patterns, or integrations that others could use, PRs are welcome.

Bugs, questions, or "I got stuck here" feedback: open an issue or run /report from within your workspace.

License

MIT - see LICENSE.