Product Manager Copilot (skills + artefacts)

▶️ New here? Watch the walkthrough on YouTube — a full tour of the skills and the setup you need to get started.

A product-management copilot built on the Pragmatic Institute Framework™ (Foundations · Focus · Build). It guides a product manager from a raw idea all the way to launch the market-driven way — and produces a client-ready artifact (Markdown or Word .docx) at every stage.

Ask it "I want to build a feature" and it won't just spit out a PRD. It routes you through the right Pragmatic stages — market problems → personas → positioning → strategy → business case → requirements → roadmap → release → launch — and hands you a finished document each step of the way.

This is an Agent Skills bundle packaged for Codex and Claude Code. The same skills/ folder works in ChatGPT desktop, Codex, Claude Code, Claude Desktop, and Cursor.

Disclaimer: This is an independent tool that operationalizes the publicly described Pragmatic Framework for practitioners. It is not affiliated with, endorsed by, or sponsored by Pragmatic Institute. "Pragmatic Framework" and "Pragmatic Institute" are trademarks of their respective owner.

https://github.com/user-attachments/assets/621169c2-778c-448c-8d8d-de8597a28fce

Contents

• What's inside
• The skills, explained
• Install
  • ChatGPT desktop / Codex
  • Claude Desktop
  • Claude Code
  • Cursor and other agents
• Artifacts: Markdown or Word
• How the journey works
• Author
• License

What's inside

1 router + 17 stage skills + a shared knowledge base.

Each skill embeds the real Pragmatic method — the urgent/pervasive/willing-to-pay test, the four market segments, the positioning-statement formula, the Priority = Market Evidence × Impact prioritization, the Strategy Matrix, opportunity scoring, roadmap-vs-release, the MRD and Release Charter templates, and more — and ends by writing the corresponding artifact.

The shared knowledge base lives in skills/pm-copilot/references/:

• framework.md — the 37-box map, definitions, rules, and reusable primitives.
• journey.md — stages → skills → artifacts, and the intent-routing table.
• artifact-output.md — the shared "ask Markdown or .docx, then write it" procedure.

The skills, explained

Each skill runs one stage of the market-driven journey and ends by writing an artifact into the feature's folder. You can invoke any of them directly, or let pm-copilot route you.

https://github.com/user-attachments/assets/520bbb41-67b0-42dd-a65c-2d52a6920fca

Router

• pm-copilot — the entry point. Figures out what you're asking for, tells you where you are in the framework, and routes you to the right stage — or runs the whole "build a feature" sequence end to end. (Orchestration; no artifact of its own.)

Foundations — understand the market

• pm-market-problems — Discover and validate real problems that are urgent, pervasive, and worth paying to solve; who to interview (the four segments), observation vs. interview, NIHITO. → Market Discovery notes + a scored Market Problems Table.
• pm-win-loss-analysis — Understand why buyers chose you or a competitor, run by an objective interviewer who evaluates the buying process, not the salespeople. → Win/Loss Interview Guide + Findings Report.
• pm-distinctive-competence — Pin down what your organization can do that rivals can't — attributes that are valued + different + provable. → Distinctive Competencies worksheet.
• pm-competitive-landscape — Map competitors and alternatives, decide when to fight and when to run, and plot the Solution Matrix. → Competitive analysis + Solution Matrix.
• pm-personas — Build evidence-based buyer and user personas (usually different people). → Buyer Persona + User Persona.
• pm-positioning — Describe the product by the problems it solves, not features; the ≤25-word positioning statement + problem-oriented capabilities. → *Positioning Document
  • POC worksheet.*
• pm-gap-analysis — Score your team across all 37 framework activities to find the biggest gaps and where to invest first. → Gap Analysis table + top-3 action plan.

Focus — turn knowledge into strategy

• pm-product-strategy — Force the who-to-win × what-to-build choice with the Strategy Matrix, and compare the team's view against leadership's. → Strategy Matrix + alignment notes.
• pm-opportunity-scoring — Rank competing opportunities objectively on competency fit, competition, investment, and buyer impact. → Opportunity Scorecard.
• pm-product-canvas — Build the business case: value prop / creation / capture, market sizing, buy-build-partner, and a go/no-go. → Product Canvas / Business Case.
• pm-product-roadmap — Communicate vision and phases as a plan, not a commitment (roadmap vs. release plan; Desire → Discover → Define → Do). → Product Roadmap.

Build — define and deliver

• pm-requirements — Turn validated problems into prioritized market requirements ("[persona] has [problem] with [frequency]", Priority = Evidence × Impact). → Market Requirements Document (MRD).
• pm-use-scenarios — Put each problem in a day-in-the-life story (Problem Card; avoid ambiguous terms; the What vs. How split). → Use Scenarios document.
• pm-prd — Assemble personas + problems + requirements + scenarios into the full spec the team builds from. → Product Requirements Document (PRD).
• pm-release-plan — Commit an increment: MVP scope, themes ("100% of something"), estimation, market window, and sign-off. → Release Plan / Charter.
• pm-stakeholder-communication — Keep everyone aligned: the Influence × Importance matrix, a comms plan, and status updates that expose risk early. → Stakeholder Communication Plan + Status Update.

Go-to-market

• pm-launch-plan — Plan a launch where readiness is more than product readiness: messaging, sales enablement, demand programs, a readiness checklist, and metrics. → Launch Plan + Readiness Checklist.

Install

ChatGPT desktop / Codex (as a personal plugin)

The Codex package is defined by .codex-plugin/plugin.json, with all 18 skills discovered from skills/.

From the ChatGPT desktop app:

1. Open Plugins and choose Add marketplace.
2. Add the GitHub repository: julianoczkowski/product-manager.
3. Open Product Manager Copilot in the marketplace and click Install.
4. Start a new task so ChatGPT loads the pm-* skills.

From the Codex CLI:

codex plugin marketplace add julianoczkowski/product-manager
codex plugin add product-manager@product-manager-marketplace

Then open or restart the ChatGPT desktop app and start a new task. To receive later updates, refresh the marketplace before reinstalling:

codex plugin marketplace upgrade product-manager-marketplace
codex plugin add product-manager@product-manager-marketplace

Claude Desktop (as a personal plugin)

Add this repo as a plugin marketplace from the app UI.

https://github.com/user-attachments/assets/d01f1a09-3438-4fc6-8dd5-687a1594f077

1. Open Settings → Customize → Connectors.
2. Next to Personal plugins, click the +, then Add → Add marketplace.
3. Choose Add from a repository.
4. Paste the repo URL — https://github.com/julianoczkowski/product-manager (or just julianoczkowski/product-manager) — and click Sync.
5. Open the Plugins → Personal directory, find Product Manager Copilot under the product-manager marketplace, open it, and click Install.

Claude Code (as a plugin)

/plugin marketplace add julianoczkowski/product-manager
/plugin install product-manager@product-manager-marketplace

Cursor and other agents (via the Skills CLI)

Install straight from this repo with Vercel's open Skills CLI:

npx skills add julianoczkowski/product-manager

Artifacts: Markdown or Word, one dossier per feature

Every stage asks how you want the deliverable — Markdown (.md) or Word (.docx).

• One dossier per feature. In hosts with full file tools (Claude Code, Cursor, VS Code Copilot) the first artifact creates a folder named after your feature/product (e.g. satellite-tracking-system/) and every artifact is saved there; in app hosts that deliver standalone files (Claude Desktop, the Codex desktop app) the filenames carry a feature-slug prefix instead. Either way, artifacts are numbered by their journey stage (06-positioning.md, 11-requirements.md) — so you accumulate a complete, ordered dossier, not scattered files.
• Sessions resume the dossier. Coming back later, the skills find the existing folder/files, reuse the header inputs, build on prior artifacts instead of re-interviewing you, and carry forward any still-open assumptions or unvalidated-evidence warnings.
• Every artifact carries the same header: Company · Feature/Product · Author · Date. These are collected once per feature and reused.
• DOCX is produced natively. The .docx is created using whatever native document-creation capability the host environment (Claude Code / Cursor / Cowork) provides — there is no bundled converter script or Python/pandoc dependency to install. Markdown stays the source of truth; if an environment can't write .docx, the skill delivers polished Markdown and tells you so rather than faking a file.

How the journey works

1. Orient — the router reads the framework map and figures out where you are.
2. Route — it maps your ask to the right stage (see journey.md).
3. Interview — the stage skill asks the few questions it needs (evidence first).
4. Draft — it produces the artifact, leading with the market problem, citing evidence, and flagging assumptions instead of inventing data.
5. Deliver — Markdown or .docx, saved into the feature's dossier (folder or slug-prefixed files, depending on the host) and numbered by journey stage so your artifacts accumulate into a complete, ordered set.
6. Advance — it names the next Pragmatic stage and offers to continue.

Author

Built by Julian Oczkowski — I build AI tools for knowledge work.

• 🎥 YouTube · @aiforwork_app — walkthroughs and AI-for-work tutorials
• ✍️ Medium — deep dives on product and AI workflows
• 💼 LinkedIn — connect and follow along

License

MIT — covers this project's own content (skill authoring, docs, manifests). The Pragmatic Framework concepts are the intellectual property of Pragmatic Institute; this project is an independent, unaffiliated practitioner aid. See NOTICE for the full trademark disclaimer and SECURITY to report issues.