PM OS Setup creates a Product Management operating system: a structured workspace with a Chief PM coordinator, specialist agents, reusable skills, product memory, and automatic indexes.
It is not just a PRD generator. It is a product mindspace your agents can actually navigate.
You give the command. The Chief PM finds the right context, routes the work, applies the right skill, creates the artifact, and keeps the workspace organized in the background.
Your taste and ways of working become reusable skills. Your docs stay linked. Your product context stays current. Your agent does not have to rediscover the map every time you ask it to help.
It works with Claude Code, Cursor, Codex, GitHub Copilot, and plain agent chats because the generated
workspace is mostly Markdown. Tool-specific wiring stays minimal by default.
- Build a structured product workspace instead of another folder of loose notes.
- Work through one Chief coordinator that routes to the right specialist automatically.
- Capture your PM taste, standards, and ways of working as reusable skills.
- Turn brainstorms, decisions, meetings, and feature ideas into durable artifacts.
- Create PRDs and decision docs without manually rebuilding context every time.
You do not need to be a developer — you just need an agent that can run skills, like Claude Code or Cursor. Don't have one yet? Install either from those links. (Cursor can also add skills from its own UI, so you can skip the terminal entirely — see Other ways to install.)
1. Install the skill:
npx skills@latest add manansuneja/pm-os-setupIf npx isn't recognized, install Node.js and try again.
2. Run it:
/pm-os-setup Set up a PM OS for Acme.
The agent asks whether to use the current folder or create a new one, then scaffolds the workspace and
offers to personalize it. If your tool doesn't show slash commands, just say it plainly:
Use pm-os-setup to set up a PM OS for Acme.
The Quickstart works in any tool with a terminal (Claude Code, Cursor, Codex, Copilot). If you don't have a terminal or would rather not use one, pick one of these instead.
If you use Claude in the desktop app or at claude.ai, upload the skill as a file:
- Download this skill as a ZIP — in GitHub, click the green Code button > Download ZIP.
- In Claude, go to Customize > Skills.
- Click +, choose + Create skill, then Upload a skill, and drop in the ZIP.
- Make sure the skill's toggle is ON in your list.
Then start a chat and ask: Use pm-os-setup to set up a PM OS for Acme.
acme-workspace/
|-- AGENTS.md
|-- CLAUDE.md
|-- INDEX.md
|-- README.md
|-- agents/
| |-- pm-chief.md
| `-- sub-agents/...
|-- product-docs/
| |-- product-vision.md
| |-- meetings/...
| |-- prds/...
| `-- decisions/...
|-- skills/
| |-- apply-pmos-struct.md
| |-- summarize-notes.md
| |-- brainstorm.md
| |-- to-prd.md
| `-- ...
`-- _setup/
|-- README.md
`-- AGENTS.md
Every PM OS content folder also includes an INDEX.md so agents can find the right context without
loading the whole workspace.
After the scaffold is created, the agent reads _setup/AGENTS.md and personalizes the workspace
around your product: product vision, Chief PM behavior, specialist agents, docs, and skills.
That personalization keeps the workspace machinery separate from the product. For example, if your product is an AI agent or robot, the product docs should describe that product's agents or robots - not accidentally turn the workspace's Chief PM into the product.
You never have to file, rename, or organize anything. Every time you create or change something — a meeting summary, a PRD, a decision, a new skill — the workspace cleans up after itself in the background: the artifact lands in the right folder with a sensible name, the indexes update so your agent can still find everything, and your product vision stays current.
So it never decays into a junk drawer of loose Markdown files. You just keep talking, deciding, and shipping — the workspace quietly keeps itself navigable around you.
On your first session the agent offers to personalize the workspace around your product, then gets out of the way. From there you just talk to the Chief PM in plain language and it routes the work:
Summarize these meeting notes and save them in the right place.
Help me brainstorm onboarding improvements and capture the best direction.
Turn this feature idea into a PRD using our product vision as context.
Update the product vision based on this new strategy note.
It comes with a handful of starter skills — summarize-notes, brainstorm, to-prd,
document-product-context — that grow with you over time (see Growing Your Skills).
The skills you start with are just a base — beginner versions meant to be reshaped around how you work. Skills are where your PM taste lives: the way you like PRDs structured, how you run a brainstorm, what a good decision doc looks like to you. The more you tune them, the more the workspace feels like an extension of you instead of a generic template.
You don't edit skill files by hand. Just tell the Chief PM what you want, in plain language:
Update the to-prd skill so PRDs always open with the problem and a one-line bet.
My brainstorms should push back harder before converging — adjust the brainstorm skill.
Create a new skill for how I like launch plans written.
Create a skill for weekly exec updates in my voice.
The Chief PM routes this through the skill-librarian, which writes or revises the skill, then keeps the workspace tidy around the change. New skills are picked up automatically — the next time the work matches, your agent applies them without being reminded.
A good habit: any time you find yourself explaining how you want something done for the second or third time, ask the Chief PM to capture it as a skill. That repeated instruction becomes reusable, and you stop re-teaching it.