Open-Source Tool Integrates PM Skills into AI Assistants

An open-source project called “PM Skills Marketplace” has been released on GitHub. It packages 68 skills and 42 chained workflows into 9 plugins that run on AI assistants, targeting product managers (PMs). The developer is phuryn. Designed for use with Claude Code and Claude Cowork, it covers the entire PM process, from discovery, strategy formulation, and execution to launch, growth, and shipping AI-generated code.

Background

While the use of AI assistants for development and data analysis is rapidly expanding, in the field of product management, there is growing demand for “structured decision support” that goes beyond simple document generation. Traditional LLMs excel at text generation but have difficulty consistently applying PM frameworks (such as Opportunity Solution Trees, hypothesis mapping, and prioritization).

PM Skills Marketplace is positioned as a tool to bridge this gap. Each skill encodes established industry frameworks such as Teresa Torres’s Opportunity Solution Tree, Marty Cagan’s product strategy, and Alberto Savoia’s experimental thinking. The design philosophy is to transform the AI assistant from a mere tool into a structured PM practice partner.

How It Works

PM Skills Marketplace is structured in three layers.

The first layer is “Skills.” Each skill provides domain knowledge, analytical frameworks, and guided workflows for specific PM tasks. A key feature is that skills are automatically loaded based on the conversational context, requiring no explicit invocation. They can also be forcibly called using the format /skill_name or /plugin_name:skill_name as needed.

The second layer is “Commands.” Commands are workflows that users invoke with /command_name. They chain multiple skills together to execute end-to-end processes. For example, the /discover command sequentially executes four skills: brainstorm-ideas → identify-assumptions → prioritize-assumptions → brainstorm-experiments.

The third layer is “Plugins.” Plugins are installable packages that group related skills and commands, categorized by PM area such as discovery, strategy, and execution. Installing this marketplace introduces all nine plugins at once.

Inter-command coordination is also a design point. After any command completes, a recommended next command is presented, allowing users to progressively advance through the PM workflow by following prompts. This design enables a complete PM process—from divergence to convergence to execution—within the AI assistant.

Installation Method

PM Skills Marketplace can be installed via two methods: Claude Cowork and Claude Code.

For Claude Cowork (recommended for non-developers), open the Customize menu at the bottom left of the Cowork screen, select Browse plugins → Personal → +. Then choose “Add marketplace from GitHub,” enter phuryn/pm-skills, and all nine plugins will be automatically installed.

For Claude Code (CLI), follow these steps:

claude plugin marketplace add phuryn/pm-skills
claude plugin install pm-toolkit@pm-skills
claude plugin install pm-product-strategy@pm-skills
claude plugin install pm-product-discovery@pm-skills
claude plugin install pm-market-research@pm-skills
claude plugin install pm-data-...

Follow the documentation for subsequent plugin installation commands. This enables commands like /discover and /strategy along with each skill.

The developer states that skills are compatible with AI assistants other than Claude Code and Cowork, but the most seamless operation is currently confirmed in the Claude Code environment.

Usage Examples

A specific use case is evaluating a new product idea. When a user enters the /discover command, Claude starts with brainstorming ideas, surfaces implicit assumptions, prioritizes them, and proposes verification experiments. The novelty is that this entire process—traditionally done manually by PMs on whiteboards or documents—can be executed interactively with an AI assistant.

Dedicated commands are also available: /write-prd for PRD (Product Requirements Document) creation, /north-star for metric definition, and /plan-launch for launch planning.

Furthermore, PM Skills Marketplace shares a design philosophy with the recently noted AI agent skill marketplaces. For example, the AI agent search “last30days-skill” covered on this site (https://singulism.com/ja/ai-agent-search-last30days-skill) can be referenced, but this project is differentiated by its specialization in the specific domain of PM work.

Editorial Comment

In the short term, PM Skills Marketplace can be considered a quick-impact tool for teams using Claude Code or Cowork. Especially in startups and small to medium-sized product teams, knowledge of PM frameworks is often not fully shared. This tool can inject best practices into such teams by “delegating processes to AI.” This could lead to changes in the quality of product discovery and the speed of decision-making over a span of 3-6 months. On the other hand, its dependence on the Claude Code environment may be a constraint to widespread adoption. How well compatibility with other AI assistants is guaranteed will become clear through practical use.

From a long-term perspective, it could influence the very definition of the PM role. If AI can handle framework application and hypothesis testing design, the required skill set for PMs is likely to shift from “memorizing frameworks” to “appropriately selecting frameworks and interpreting results.” Over a span of 1-3 years, “PM skills premised on AI assistants” may become standardized in the product management domain, altering education and hiring criteria. However, whether such tools truly increase the probability of product success remains to be verified.

The editorial team asks: will the automated application of frameworks improve decision quality, or will it induce shallow thinking as a mental shortcut? Furthermore, how this skill marketplace aggregates and updates PM community knowledge, and what moderation mechanisms will be implemented, are also noteworthy challenges for the future.

Reference

• phuryn/pm-skills - GitHub — Published June 11, 2026