`phuryn/pm-skills`: 68 PM frameworks wired into your agent — including a new red-team mode

Most agents write fast but decide slowly. They'll draft a PRD or sprint plan without pausing to ask whether the underlying assumptions actually hold. phuryn/pm-skills is a 68-skill, 9-plugin marketplace that routes around that — it encodes established PM frameworks from Teresa Torres, Marty Cagan, and Alberto Savoia directly into your agent's workflow, so the discipline runs automatically instead of relying on you to prompt for it. 1

v2.0.0 shipped three days ago (June 5, 2026) with two additions worth installing for: a /red-team-prd command that attacks your plan's assumptions before the market does, and a new pm-ai-shipping plugin that gives vibe-coded apps a reviewer-ready audit trail. 2 The project has 12,500+ GitHub stars and 70,600+ total installs across skills.sh as of today. 1 3

What's inside

The marketplace is organized into three layers. Skills are domain-knowledge encapsulations — Teresa Torres' Opportunity Solution Tree, Ash Maurya's Lean Canvas, Porter's Five Forces — that load automatically when the topic matches, with no explicit invocation needed. Commands are slash-triggered workflows that chain multiple skills end-to-end. Plugins group related skills and commands into installable units per PM discipline. 1

The nine plugins cover the full product lifecycle:

All 9 plugins at a glance: Product Discovery, Strategy, Execution, Market Research, Data Analytics, Go-to-Market, Marketing & Growth, Toolkit, and the new AI Shipping Kit — All 9 plugins at a glance — the Execution plugin picked up the `strategy-red-team` skill in v2.0, and AI Shipping Kit is new. 1

Five slash commands carry most of the weight day-to-day: /discover (product discovery loop), /strategy (strategy canvas), /write-prd (PRD generation), /plan-launch (GTM strategy), and /north-star (North Star metric definition). 1

As the README puts it: "Each skill encodes a proven PM framework — discovery, assumption mapping, prioritization, strategy — and walks you through it step by step. You get the rigor of Teresa Torres, Marty Cagan, and Alberto Savoia built into your daily workflow, not sitting on a bookshelf." 1

Install

Claude Cowork (no CLI required): Open Customize (bottom-left) → Browse plugins → Personal → + → Add marketplace from GitHub → type phuryn/pm-skills. All 9 plugins install at once. 1

Note: two open issues report that this path fails for some users. Issue #7 (filed March 2026) documents a "Failed to add marketplace" error in Claude Desktop v1.1.5749, and issue #26 notes the Cowork UI changed after the April 24, 2026 Claude update — users on the $20/mo plan may no longer see the Personal skills filter. 4 5 If the UI path fails, use the CLI route below.

Claude Code (CLI):

claude plugin marketplace add phuryn/pm-skills

# Then install each plugin individually:
claude plugin install pm-toolkit@pm-skills
claude plugin install pm-product-discovery@pm-skills
claude plugin install pm-product-strategy@pm-skills
claude plugin install pm-execution@pm-skills
claude plugin install pm-market-research@pm-skills
claude plugin install pm-data-analytics@pm-skills
claude plugin install pm-go-to-market@pm-skills
claude plugin install pm-marketing-growth@pm-skills
claude plugin install pm-ai-shipping@pm-skills

Any other agent (Cursor, Gemini CLI, OpenCode, Kiro):

npx skills add phuryn/pm-skills

This installs all 68 skill files. Slash commands (/discover, /write-prd, etc.) are Claude-only — on Codex or Cursor you describe the workflow in plain language and the skills fire contextually. 1

github.com · GitHub 仓库

phuryn/pm-skills

https://github.com/phuryn/pm-skills

正在加载内容卡片…

The two v2.0 additions worth your attention

`/red-team-prd` — adversarial assumption pressure-test

This is the command that caught attention when v2.0 dropped. Feed it a plan, roadmap line, or PRD excerpt and it returns the assumptions behind each claim, the conditions under which each assumption fails, the evidence needed to validate it, and the cheapest test to get that evidence first.

Paweł Huryn's framing for the command: "The impolite questions, before the market asks them." And on what it won't do: "A useful red team does not manufacture doubt to look smart. If the plan is strong, it should say so." 6

A concrete example from the documentation: input /red-team-prd Prioritize AI onboarding, activation is our bottleneck. Output: "Claim: Activation is the constraint on growth. Fails if: retention, not activation, is where users actually drop off. Cheapest test: pull the activation-to-retention funnel for the last 3 cohorts. One afternoon, no engineering." 6

The point, in Huryn's words: "to kill a bad bet on a Tuesday instead of defending it in front of leadership a quarter later."

This sits inside pm-execution alongside the pre-existing /pre-mortem command. The distinction matters: /pre-mortem imagines the plan has already failed and narrates backward; /red-team-prd attacks live assumptions while there's still time to run a test. 6

`pm-ai-shipping` — making vibe-coded apps auditable

The ninth plugin addresses a specific gap that AI-assisted development has created: agents write code fast but leave no record of intent. There's no document describing what the system should do, who can do what, where secrets live, or which rules are actually enforced. 6

The plugin ships two skills and five commands:

shipping-artifacts defines the persistent documentation set an AI-built app needs to be reviewable — architecture overview, user and permission flows, variable/secret map, test-coverage map
intended-vs-implemented audits the gap between what the documentation promises and what the code actually does

The five commands — /document-app, /security-audit-static, /performance-audit-static, /derive-tests, and /ship-check — can be run by a non-technical PM. The gate the plugin checks is straightforward: do the permissions documented match what the code enforces? Do the tests actually cover the flows described? 6

Three example prompts: tailoring a resume (skill), writing a PRD with /write-prd (command), and auditing an AI-built app with /ship-check (command) — Examples of a skill invoked by context and two slash commands in action — all running on Claude with Opus 4.8. 1

Community reception

The project has gone through 1,500+ forks in the three months since its March 2026 launch, with localization forks in Korean, Turkish, and Persian. 7 A post by @polydao in April 2026 calling out pm-skills as part of a "Top 67 Claude Skills that turn a $20 subscription into a full dev team" gathered 951,493 views. 8

Jeff Gothelf (author of Lean UX) praised the technical implementation on LinkedIn while flagging the structural limitation clearly: "You can encode a framework into a prompt chain. You can't encode the wisdom of knowing when to and not to use it." His specific concern: "When a PM types /discover and Claude chains four skills automatically — the workflow runs. The output looks awesome. But that PM just skipped the hard part which is deciding which discovery question matters most right now." 9

On skills.sh's independent platform, the test-scenarios skill carries 4.8 stars from 39 reviews. 10 The closest competitor on the same platform, product-on-purpose/pm-skills (Jonathan Prisant), has 11,100 installs vs. phuryn's 70,600 — though the competitor's repo has more governance infrastructure: a SECURITY.md, CODE_OF_CONDUCT.md, CHANGELOG.md, and CI-enforced contracts that phuryn/pm-skills currently lacks. 11 3

When NOT to use this

You need one framework, not 68. If your current problem is "write a better PRD" and nothing else, the full marketplace is overhead. Install the pm-execution plugin or just use npx skills add phuryn/pm-skills --skill create-prd to get the single skill without loading everything. The 68 skills are "loaded automatically when relevant" — all their description fields sit in your agent's context simultaneously, which adds token overhead that hasn't been publicly measured. 1

You're a first-time PM learning the frameworks. The skills assume you can evaluate whether the output is correct. Jeff Gothelf's point cuts directly here: /discover chains brainstorming → assumption identification → prioritization → experiment design. If you don't already understand why that sequence matters — what makes an assumption "high risk," what distinguishes an experiment from a feature — the output gives you documents without the judgment to act on them. 9

You need guaranteed long-term maintenance. Paweł Huryn is a single maintainer running five-plus other open-source projects alongside a 131,000-subscriber newsletter. 12 Ten issues have been open since March 2026 with zero triage — no assignee, no label, no maintainer reply on the two confirmed installation bugs (#7 and #26). 13 For a prototype or personal workflow, that's fine. For a team dependency, fork it. There's no SECURITY.md and no public security audit covering the 68 skill files. 11

The bottom line

The real value in phuryn/pm-skills is the opinionated sequencing. /discover doesn't just brainstorm ideas — it forces assumption mapping and experiment design before you can move on. /red-team-prd asks the adversarial questions before a quarter of engineering time is spent defending a weak premise. For a solo builder or a small team shipping fast, that pre-built discipline is worth more than the individual outputs.

The bus factor of 1 and the unresolved installation bugs are real considerations. But for a daily workflow tool on a $20 Claude subscription, this hits the target: you get Teresa Torres and Marty Cagan running alongside your agent, not sitting on a bookshelf.

Cover image: AI-generated illustration