PM Skills 2.0 Is Half a Funeral and Half a Blueprint

Paweł Huryn just shipped the most useful PM tooling on GitHub, and half of it is embalming a role I have spent a year arguing should die. PM Skills 2.0 is 68 Claude skills, 42 commands, and 9 plugins, and the verdict on it splits cleanly down the middle. Both things are true, and the interesting part is figuring out which half you should install.

The short version

PM Skills 2.0 by Paweł Huryn is the best free PM tooling on GitHub: 68 Claude skills, 42 commands, 9 plugins, MIT-licensed, roughly 12,000 stars. Two pieces are new thinking. The /red-team-prd command turns any PRD or roadmap into ranked kill-assumptions with fails-if conditions and cheap tests. The AI Shipping Kit (/ship-check) audits the gap between what an AI-built system was supposed to do and what the code actually does, which is the unsolved accountability problem of 2026. The rest of the catalog encodes the legacy PM artifact stack, RICE matrices and PRD templates and canvases, and automating that scaffolding just builds scaffolding at machine speed. Install pm-ai-shipping and pm-execution, skip the other seven plugins, and do not mistake faster documents for better judgment.

Let me give credit first, because it is earned.

What Huryn got right

PM Skills 2.0 is 68 Claude skills, 42 commands, and 9 plugins. Free, MIT-licensed, 12,000 GitHub stars. While most PM influencers were posting carousels about AI, Huryn built working infrastructure and gave it away. That is Product Builder behavior, the kind I argued for in Cagan Built the Church. I Argue We Don't Need One., and the irony that it comes from someone packaging classic PM frameworks is not lost on me.

Two pieces of this release are new thinking.

/red-team-prd takes your roadmap, PRD, or one-line bet and returns the assumptions ranked by what hurts most, what is most likely wrong, and what is cheapest to test this week. Each claim comes with a fails-if condition and a concrete test. Five kill-assumptions with tests instead of twenty generic risks. This is discovery compressed from three weeks of ritual into thirty seconds of confrontation, and it is exactly the direction the discipline needs to move. It is assumption testing with the ceremony stripped out.

The AI Shipping Kit is the bigger deal, and almost nobody will say so. The /ship-check command documents an AI-built repo, audits the gap between intended and implemented, maps test coverage, and compiles a packet a human can sign off on. The unenforced permission. The scheduled-job endpoint anyone can call. The test that proves nothing.

This is the unsolved problem of 2026. Agents made building cheap and made accountability expensive. Everyone is shipping AI-written code, and almost nobody has a written record of what the system was supposed to do. Huryn built the missing artifact layer. I am building an agentic healthcare product, and intended-vs-implemented is not a nice-to-have in my world. In regulated software it is the entire job. It is the same instinct behind The Eval-First Product Org: the quality record is the product now.

So that is the credit. Now the knife.

What the catalog gets wrong

Huryn ends his post with a line I could have written: AI does not fix weak product judgment, it scales it.

Correct. Now apply that sentence to his own catalog.

A meaningful chunk of those 68 skills encodes the legacy PM stack: prioritization frameworks, PRD templates, discovery ceremonies, and stakeholder artifact generation. Encoding RICE into a skill does not make RICE good. It makes RICE faster. You have taken a framework whose main output was the appearance of rigor and given it an engine.

That is the trap with this entire category of tooling. The frameworks were never the judgment. They were scaffolding for people who had not yet developed judgment, and in too many organizations they became a substitute for it. Automating the scaffolding does not build the building. It builds scaffolding at machine speed. Judgment comes from reps(coming Jul 13), not from templates that render faster.

I can already see the failure mode, because I have watched it happen inside companies. A mid-level PM installs all 9 plugins, and their weekly output triples. More PRDs, more prioritization matrices, more beautifully structured strategy docs. Their actual contact with customers and code does not change at all. They have become a more prolific version of exactly what they were. The feature factory did not close. It got a better printer. This is the same dead job I found when I reviewed 200 PM job descriptions, now with better tooling.

There is a second problem: catalog bloat is a product smell. 68 skills and 42 commands is a surface area no PM needs. Huryn's own defense is that you do not have to learn the catalog because the right skill surfaces automatically. Maybe. But when the honest pitch is that most of your product invokes itself invisibly, the real product is about five commands wearing a 68-skill costume.

And the deepest tension: the two best pieces of this release quietly contradict the rest of it. /red-team-prd exists to kill plans before documents calcify around them. /ship-check exists because the unit of truth is now the repo, not the spec. Both point toward a world where the artifact stack the other 60 skills generate matters less every quarter. Huryn built the tools of the next era and bundled them with a museum of the last one.

To be fair, that may be deliberate. You meet 12,000 PMs where they are, not where you wish they were. As distribution strategy it is smart. As a philosophy it is a bridge, and bridges are for crossing, not living on.

What you should actually do with it

Direct answers, because that is the question.

Install two plugins, not nine. pm-ai-shipping and pm-execution. The Shipping Kit and the red-team command are the load-bearing pieces. Treat the rest as optional reference, not workflow.

Run /red-team-prd on your own bets before anyone else sees them. Not as a checkbox. Read the fails-if conditions and actually pull the funnel it tells you to pull. The command kills a bad bet on a Tuesday. Only you can decide to let it.

Learn /ship-check like your title depends on it, because it does. If you are signing off on AI-built features and you cannot articulate intended versus implemented, you are not doing product management. You are doing product hoping. This workflow, or something like it, is becoming the core competence of anyone who ships with agents. It is the closest thing in this library to a job description for the Product Builder.

Do not install your judgment. If you find yourself running /write-prd to produce a document nobody asked for, stop and ask what decision the document was supposed to force. Then make the decision. The skill that matters most is not in the marketplace. If you want skills that do encode your own practice instead of someone else's framework, build them yourself.

Steal the meta-lesson. Huryn did not write a hot take about AI and PMs. He shipped a repo, versioned it, and let 12,000 people stress-test it. Whatever you think of the contents, that is the correct way to have a product opinion in 2026. The artifact is the argument.

Your first hour with it

If you want the forward-pointing half installed before lunch, here is the exact sequence.

In Cowork: open Customize, go to Browse plugins, then Personal, then add marketplace from GitHub and enter phuryn/pm-skills. In Claude Code:

claude plugin marketplace add phuryn/pm-skills
claude plugin install pm-ai-shipping@pm-skills
claude plugin install pm-execution@pm-skills

Then run two tests on your real work, not a demo. First, paste your current top bet into /red-team-prd and pull the single cheapest data check it surfaces before the end of the day. Second, point /ship-check at the last AI-built feature you signed off on and read the intended-vs-implemented gaps. If either output changes a decision this week, the install paid for itself. If neither does, you learned something about your bets instead.

One warning from running my own agent fleet: tools you install but never wire into a weekly cadence die quietly. Put /red-team-prd in your bet-writing ritual and /ship-check in your pre-launch checklist, or delete them.

PM Skills 2.0 is the best free tooling the PM profession has produced, built by someone with real builder instincts, and roughly half of it accelerates a way of working that is on its way out. Take the half that points forward.

The repo is at github.com/phuryn/pm-skills. Go form your own opinion. With a prototype, obviously.

Sources: PM Skills Marketplace on GitHub, Paweł Huryn's PM Skills 2.0 announcement on The Product Compass