What You Can Learn from the pm-skills README: A Complete Guide to the AI Operating System for Product Managers

The pm-skills README explains how to transform Claude, Codex, and other AI assistants into specialized product management tools through a modular system of skills, slash commands, and plugins that cover the entire product lifecycle.

The phuryn/pm-skills repository serves as an AI operating system that encodes product management frameworks into machine-readable formats. According to the source code, the README documents the architecture, installation methods, and nine plugin categories that equip AI models with domain-specific capabilities ranging from opportunity solution trees to go-to-market strategies.

Architecture: Skills, Commands, and Plugins

The README outlines a three-layer architecture that structures how AI assistants access product management knowledge.

Skills as Knowledge Modules

Skills are self-contained markdown files stored at */skills/*/SKILL.md that encode specific frameworks. Each skill represents a discrete product management capability—such as SWOT analysis, pricing strategy, or user persona creation—that the AI can invoke automatically when it detects relevance. Users can force a specific skill using the /plugin-name:skill-name prefix syntax.

Commands as Workflow Orchestrators

Commands function as slash-style workflows declared in */commands/*.md files. These chain multiple skills into end-to-end processes. For example, the /discover command runs a sequence of discovery skills including "brainstorm-ideas-new," "identify-assumptions-new," and "prioritize-assumptions" to guide users through product discovery.

Plugins as Domain Bundles

Plugins are logical groupings of related skills and commands. The README defines nine plugins in the marketplace, each targeting a specific stage of the product lifecycle. Installing a plugin brings the entire suite for that domain into the AI's context window.

Installation Options

The README provides three distinct installation paths depending on your technical environment and AI assistant of choice.

Claude Cowork (Recommended for Non-Developers)

1. Open Customize in Claude
2. Select Browse plugins
3. Click Add marketplace from GitHub
4. Enter phuryn/pm-skills

This method delivers all nine plugins with ready-to-use slash commands through the graphical interface.

Claude Code (CLI)

For terminal-based workflows, use the following commands:

# Add the marketplace registry

claude plugin marketplace add phuryn/pm-skills

# Install specific plugins

claude plugin install pm-product-discovery@pm-skills
claude plugin install pm-execution@pm-skills

Codex CLI (OpenAI)

When using OpenAI's Codex, skills are available for prompting but commands are not exposed as slash commands:

codex plugin marketplace add phuryn/pm-skills
codex plugin add pm-toolkit@pm-skills

Available Plugins and Capabilities

The README enumerates nine plugins covering the full product management spectrum. Each plugin contains specific skills and corresponding slash commands.

Plugin	Core Purpose	Example Skill	Example Command
pm-product-discovery	Ideation through experiment design	opportunity-solution-tree	/discover
pm-product-strategy	Vision and business model development	product-vision	/strategy
pm-execution	PRDs, OKRs, and sprint planning	create-prd	/write-prd
pm-market-research	Personas and competitor analysis	user-personas	/research-users
pm-data-analytics	SQL generation and cohort analysis	sql-queries	/write-query
pm-go-to-market	Beachhead strategy and battlecards	gtm-strategy	/plan-launch
pm-marketing-growth	Positioning and North Star metrics	north-star-metric	/north-star
pm-toolkit	Resume review and legal documents	review-resume	/review-resume
pm-ai-shipping	Documentation for AI-generated code	shipping-artifacts	/ship-check

These definitions reside in the plugin-specific README files such as pm-toolkit/README.md and pm-execution/README.md.

How to Use Slash Commands

Commands are invoked with a leading forward slash. The README demonstrates concrete invocation patterns:

/discover AI-powered meeting summarizer for remote teams
/strategy B2B project management tool for agencies
/write-prd Smart notification system that reduces alert fatigue
/north-star Two-sided marketplace connecting freelancers with clients

Each command executes its skill chain and then suggests the next logical command, creating a guided workflow from discovery through execution.

Extending the System

The README emphasizes portability and extensibility. Skill files follow a universal markdown format that can be copied into other AI assistants including Gemini, OpenCode, Cursor, and Kiro.

To add new capabilities:

1. Create a SKILL.md file under the appropriate plugin's skills/ directory
2. Or create a new *.md file under commands/ to define slash workflows
3. The marketplace automatically indexes new files without requiring configuration updates

Summary

• The pm-skills README documents an AI operating system that turns general-purpose models into product management specialists through structured markdown files.
• The architecture separates skills (individual frameworks), commands (workflow chains), and plugins (domain-specific bundles).
• Installation supports Claude Cowork, Claude Code CLI, and Codex CLI, with nine plugins covering discovery, strategy, execution, and go-to-market phases.
• Skills are automatically loaded when relevant, or forced via /plugin-name:skill-name syntax.
• The system is fully extensible by adding markdown files to */skills/ and */commands/ directories.

Frequently Asked Questions

What file format do pm-skills use for knowledge storage?

The system uses markdown files with specific naming conventions. Skills are stored as SKILL.md within plugin subdirectories (e.g., pm-product-discovery/skills/), while commands are defined as *.md files within */commands/ folders. This plain-text approach ensures compatibility across Claude, Codex, Gemini, and other AI assistants.

How does the automatic skill detection work?

According to the README source, the AI model automatically loads relevant skills when it detects contextual cues in the conversation. Users can override this behavior and force specific skills using the prefix syntax /plugin-name:skill-name, such as /pm-product-discovery:opportunity-solution-tree, ensuring precise control over which frameworks the AI applies.

Can I use pm-skills without Claude?

Yes. While slash commands require Claude's Cowork or Code interfaces, the skill markdown files can be manually copied into prompts for other AI assistants including OpenAI's Codex, Google's Gemini, Cursor, and OpenCode. The README includes a dedicated section "Other AI assistants (skills only)" explaining this portable approach.

What is the difference between a skill and a command?

A skill is a single knowledge module that teaches the AI one specific framework (like creating user personas). A command is a workflow that chains multiple skills together—for instance, /discover orchestrates ideation, assumption mapping, and experiment design skills into a complete discovery process.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}