# phuryn/pm-skills: a Claude Code marketplace of PM frameworks as installable plugins > The repository packages 69 product management skills and 42 chained workflows into 9 plugins for Claude Code, Cowork and Codex CLI. It is a framework library, not an assistant, and its limits show up the moment you leave Claude. **phuryn/pm-skills** — PM Skills Marketplace: 100+ agentic skills, commands, and plugins — from discovery to strategy, execution, launch, and growth. - Repository: https://github.com/phuryn/pm-skills - Website: https://www.productcompass.pm/p/pm-skills-2-red-team-ship - Stars: 26,667 · Forks: 2,838 - Language: Unknown - License: MIT - Published: 2026-09-21 · Updated: 2026-09-21 · Language: en - Canonical page: https://hysenlabs.com/projects/phuryn-pm-skills ## The problem phuryn/pm-skills addresses: generic model output versus a named PM framework Ask an assistant to write a PRD and you get a PRD-shaped document. Ask it to run discovery and you get a list of ideas with no stated method behind them. The README frames the gap directly: "Generic AI gives you text. PM Skills Marketplace gives you structure." The structure is the product. Each skill in the repository encodes a specific product management framework, and the repository attributes the lineage to Teresa Torres, Marty Cagan and Alberto Savoia. The intended user is a product manager who already knows what discovery, assumption mapping and prioritisation are, and wants the assistant to walk through those steps rather than summarise them. It is not a research tool and it does not gather data. It supplies sequence and prompts. The README's Start Here list maps a situation to a command: a new idea goes to /discover, strategic clarity to /strategy, a PRD to /write-prd, launch planning to /plan-launch, metric definition to /north-star. That mapping is the clearest statement of scope in the repository. ## Skills, commands and plugins: how the three layers fit together The repository uses three distinct objects, and conflating them causes most of the confusion around it. Skills are the base layer. A skill gives the assistant domain knowledge or a guided workflow for one PM task. The README states that skills load automatically when the conversation makes them relevant, with no explicit invocation. When you need a skill to win over the model's general knowledge, the README gives two forcing forms: /plugin-name:skill-name or /skill-name, with the note that Claude adds the prefix. Commands are user-triggered. They are invoked as /command-name and chain one or more skills. The README's worked example is /discover, which chains four skills in order: brainstorm-ideas, then identify-assumptions, then prioritize-assumptions, then brainstorm-experiments. That chain is the architecture in miniature: the command is the orchestration, the skills are the steps. Plugins are the installable unit. Nine of them exist, each covering a PM domain, and installing the marketplace brings all nine in at once. The README notes that some skills, naming prioritization-frameworks and opportunity-solution-tree, are standalone references the model draws on whenever relevant, with no command attached. Commands also suggest the next command when one finishes, which is how the workflows are meant to be followed end to end. ## Installing pm-skills in Claude Code and running /discover for the first time Claude Cowork is the path the README recommends for non-developers, and it is entirely graphical: open Customize in the bottom-left, go to Browse plugins, then Personal, then the + control, choose Add marketplace from GitHub, and enter phuryn/pm-skills. All nine plugins install together. For Claude Code, installation is two steps. The first registers the marketplace, the second installs plugins one at a time: ```bash claude plugin marketplace add phuryn/pm-skills claude plugin install pm-toolkit@pm-skills claude plugin install pm-product-discovery@pm-skills ``` The README lists nine plugins in total, covering pm-toolkit, pm-product-strategy, pm-product-discovery, pm-market-research, pm-data-analytics, pm-marketing-growth, pm-go-to-market, pm-execution and pm-ai-shipping. Installing all nine means running the install line once per plugin. The README advises installing whole plugins rather than picking individual skills, on the grounds that a workflow usually relies on several skills that ship together. Once installed, a first real use is the discovery chain. Type /discover in Claude Code and the command runs brainstorm-ideas, identify-assumptions, prioritize-assumptions and brainstorm-experiments in sequence. What you should see is a staged conversation that pauses at each step rather than a single wall of text. After it completes, the README states the command suggests relevant next commands. ## Codex CLI and other assistants: what survives the move The README makes a specific and slightly unusual claim: Codex reads the same plugin marketplace file as Claude Code, so installation needs no conversion or file copying. The commands mirror the Claude Code ones with codex substituted for claude: ```bash codex plugin marketplace add phuryn/pm-skills codex plugin add pm-product-discovery@pm-skills ``` The catch is stated plainly. The slash commands install but do not run, because Codex plugins do not expose commands. What you get is the skills, invocable by name. To run a workflow you describe the steps in plain language. The README's own example asks Codex to run product discovery on an idea, listing the four steps and ending with the instruction to pause between each step. There is a second option the README describes as best-effort. Because the command files ship inside each installed plugin, you can ask the model to read them and generate equivalent Codex skills. The README warns that some Claude-specific command syntax will not translate. Treat that as a conversion you inspect rather than trust. For Gemini CLI, OpenCode, Cursor and Kiro, only the skills travel. The README's table says to copy skill folders into .gemini/skills/, .opencode/skills/, .cursor/skills/ or .kiro/skills/ respectively, and marks every one of them as skills only. The repository layout supports this: the skills live as skills/*/SKILL.md files following what the README calls the universal skill format. A shell snippet in the README shows a loop over pm-*/ directories creating .opencode/skills/, though the excerpt is truncated mid-command. ## Where pm-skills stops being the right tool The largest constraint is portability of the workflow layer. Skills move; commands do not. If your team standardises on Codex CLI, Cursor or Gemini CLI, you inherit a set of framework documents and lose the chained orchestration that the README treats as the main value. You can rebuild that orchestration by hand, but you are then maintaining a parallel copy that the upstream repository will not track. The second constraint is that the repository is a prompt and framework library, not a system of record. Nothing in the README describes persistence of outputs, versioning of the artefacts a command produces, or any integration with a tracker, a docs tool or a data warehouse. If you need the output of /write-prd to land somewhere specific, that plumbing is yours to build. The third is selection. The README advises installing whole plugins, which cuts against the instinct to take one skill and leave the rest. Following that advice means accepting a larger surface than you asked for, including skills you may never invoke. The README does not document an uninstall or rollback procedure, so a team experimenting with the marketplace should decide in advance how it will back the installation out. Finally, the frameworks are opinionated by design. If your organisation runs a different discovery method, the chained commands will walk you through steps that do not match your process, and the automatic skill loading means those frameworks can enter a conversation you did not intend to route through them. ## How pm-skills differs from a general-purpose agent skill collection The obvious alternative is assembling your own skills directory: write SKILL.md files for the frameworks your team actually uses and place them in .claude/skills/ or the equivalent path for your assistant. That approach gives you exactly the frameworks you want and nothing else, and it keeps working if the upstream project changes direction. The difference is in the chaining. A hand-written skills directory gives the model reference material it may draw on. pm-skills adds a command layer on top, where /discover is a fixed four-step sequence rather than a set of documents the model might consult. That ordering is the part you cannot get from a folder of markdown, and it is also the part that locks you to Claude Code or Cowork. A second alternative is to keep using the assistant with no skills at all and supply the framework in the prompt each time. That works for a one-off PRD. It degrades when several people on the same team need to produce comparable artefacts, because the framework then lives in whoever wrote the prompt rather than in the repository. The honest comparison is between a fixed, opinionated chain that installs in two commands and a bespoke set of skills that costs you an afternoon to write but matches your process exactly. pm-skills is the faster start and the weaker fit. ## Maintenance, upgrade cost and the MIT licence The repository is not archived, and the last push was on 2026-09-14. Two releases are listed: v2.0.0 on 2026-06-05 and v2.1.0 on 2026-07-03. The README's own framing of the project as an operating system for product decisions, combined with a v2 line, means the plugin contents have changed across major versions rather than accumulating quietly. Upgrade cost depends on how much of the marketplace you installed. Because commands chain skills, a change to one skill can alter the behaviour of every command that references it, and the README explicitly notes that some skills serve multiple commands. Upgrading a single plugin in isolation is therefore not obviously safe. The repository ships a validate_plugins.py script and a tests/ directory at the top level, plus a GitHub Actions workflow referenced by the README's Tests badge, which suggests plugin structure is checked in CI. That is a signal about internal consistency, not about whether an upgrade will change what a command outputs. The licence is MIT, declared in the LICENSE file and shown as a badge in the README. MIT permits commercial use, modification and redistribution provided the copyright notice and permission notice are preserved. For a team embedding these skills in an internal workflow, the practical implication is that you can fork and adapt them. This is a description of the licence text, not legal advice, and the terms that matter to you should be read in the LICENSE file itself. ## Conclusion Adopt it if your product team already works inside Claude Code or Cowork and wants discovery, prioritisation and launch work to follow a repeatable structure rather than a blank prompt. Skip it if you use Codex CLI as your only assistant and expect /discover or /write-prd to behave as slash commands, or if you want a single skill rather than a plugin, since the README advises installing whole plugins because workflows depend on several skills shipping together. Before rolling it out, verify two things: that your Claude Code version accepts claude plugin marketplace add, and that the plugin you intend to use actually contains the command you plan to run, because the README does not document a rollback path after installation. ## FAQ ### What are some PM skills in phuryn/pm-skills? The repository groups its skills into nine plugins covering toolkits, product strategy, product discovery, market research, data analytics, marketing growth, go-to-market, execution and AI shipping. Individual skills named in the README include prioritization-frameworks and opportunity-solution-tree, and the /discover command chains brainstorm-ideas, identify-assumptions, prioritize-assumptions and brainstorm-experiments. ### What is pm skills? pm skills is phuryn/pm-skills, a marketplace of 69 product management skills and 42 chained workflows grouped into 9 installable plugins. It is designed for Claude Code and Cowork, with the skills compatible with other AI assistants. ### What are the top 5 skills of a project manager in this marketplace? The README does not rank skills by importance, but its Start Here list points to five entry points: /discover for a new idea, /strategy for strategic clarity, /write-prd for a PRD, /plan-launch for launch planning and /north-star for defining metrics. ## Sources - [License: MIT](https://github.com/phuryn/pm-skills/blob/main/LICENSE) - [phuryn/pm-skills on GitHub](https://github.com/phuryn/pm-skills) - [Project website](https://www.productcompass.pm/p/pm-skills-2-red-team-ship) - [README](https://github.com/phuryn/pm-skills/blob/main/README.md) - [Releases](https://github.com/phuryn/pm-skills/releases) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/phuryn-pm-skills