What Are PM Skills and How Do They Work: A Technical Guide to the phuryn/pm-skills Framework

PM Skills are a curated collection of markdown-based frameworks that turn Claude into a product management assistant through 9 specialized plugins containing 68 skills and 42 slash commands.

PM Skills are a structured knowledge base hosted in the phuryn/pm-skills repository that encode proven product management methodologies into machine-readable formats. Unlike generic AI prompting, these skills provide domain-specific frameworks for discovery, strategy, execution, and analytics that load automatically when Claude detects relevant requests or when users invoke specific slash commands.

Core Architecture: Skills, Commands, and Plugins

The phuryn/pm-skills framework organizes product management knowledge into three hierarchical concepts that work together to deliver structured workflows.

Skills

A Skill is a self-contained markdown file (SKILL.md) that provides domain knowledge, analytical frameworks, or guided workflows for a single PM task. Skills live inside plugin-specific directories and follow a universal schema that Claude parses automatically. For example, pm-product-discovery/skills/brainstorm-ideas-existing/SKILL.md contains the methodology for generating ideas from existing products, including prompting logic and output formats.

Skills trigger automatically when the model detects relevant request phrases (like "riskiest assumptions") or through explicit invocation using the /plugin-name:skill-name prefix.

Commands

A Command is a slash-command (e.g., /discover) that chains one or more skills into an end-to-end workflow. Commands are user-triggered and map directly to real-world PM processes. The command definitions reside in markdown files within each plugin's commands/ directory. For instance, pm-product-discovery/commands/discover.md defines the /discover command that runs ideation, assumption mapping, prioritization, and experiment design sequentially.

Plugins

A Plugin is a top-level folder (e.g., pm-product-discovery, pm-execution) that bundles a thematic set of skills and commands. The repository contains 9 independent plugins that group related functionalities:

• pm-product-discovery: Ideation, assumption mapping, and user research
• pm-strategy: Product strategy canvases and roadmap planning
• pm-execution: PRD writing, sprint planning, and delivery workflows
• pm-market-research: Competitive analysis and market sizing
• pm-data-analytics: SQL generation, cohort analysis, and metric definitions
• pm-go-to-market: Launch planning and GTM strategy
• pm-marketing-growth: Growth experiments and marketing frameworks
• pm-ai-shipping: AI-specific shipping workflows and documentation
• pm-utilities: Cross-functional tools and templates

According to CLAUDE.md, the marketplace comprises 68 skills and 42 commands across these nine plugins.

How PM Skills Work: The Execution Flow

Understanding the runtime behavior of PM Skills requires following the sequence from installation to artifact delivery.

Installation and Loading

Users add the marketplace via Claude Cowork, Claude Code CLI, or compatible agents. The system pulls all nine plugins simultaneously, making skills and commands available immediately. Skills are not loaded into context until triggered, preserving token efficiency.

Skill Detection and Loading

As conversations progress, Claude scans user prompts for trigger phrases defined in skill files. When a match occurs (e.g., "brainstorm ideas" triggering brainstorm-ideas-existing), the corresponding SKILL.md file injects into the model's context window. Users can bypass auto-detection by prefixing requests with /plugin-name:skill-name.

Command Invocation and Orchestration

When a user types a slash command like /discover AI-powered meeting summarizer, the command dispatcher loads the ordered skill list defined in the command's markdown file. Each skill executes sequentially, prompting for inputs or producing intermediate outputs before chaining to the next step. The pm-product-discovery/commands/discover.md file illustrates this declarative orchestration pattern.

Result Delivery and Chaining

After the final skill completes, the system returns a structured artifact—such as a product discovery report, PRD, or cohort analysis spreadsheet. The framework then suggests "next-step" commands based on the artifact type, creating a natural workflow progression from discovery to execution.

Practical Usage Examples

The following commands demonstrate real-world usage patterns within the phuryn/pm-skills ecosystem:

Command	Input	Output
/discover	/discover AI-powered meeting summarizer for remote teams	Complete discovery cycle: brainstorming, assumption mapping, prioritization matrix, and experiment proposals
/strategy	/strategy B2B project-management tool for agencies	Nine-section Product Strategy Canvas (vision, target market, value proposition, monetization)
/write-prd	/write-prd Smart notification system that reduces alert fatigue	Eight-section PRD ready for engineering hand-off, generated via pm-execution/skills/create-prd/SKILL.md
/analyze-cohorts	/analyze-cohorts Weekly retention for users who signed up in January vs February	Retention curves and statistical insights
/north-star	/north-star Two-sided marketplace connecting freelancers with clients	North Star Metric definition with supporting input metrics

For direct skill queries without commands, users can simply ask: "What are the riskiest assumptions for our AI writing assistant idea?" This triggers the prioritize-assumptions skill directly via the pm-product-discovery plugin.

Extending the Framework

The markdown-based architecture enables zero-code customization.

Adding New Skills

Create a new SKILL.md file in any plugin's skills/ directory following the universal schema. The file becomes instantly discoverable because Claude loads any markdown file in these directories. Skills like prioritization-frameworks demonstrate cross-plugin reuse, residing in one location but callable from multiple commands.

Adding New Commands

Define new workflows by creating markdown files in a plugin's commands/ folder. Commands are declarative—simply list the ordered skills to execute. No compilation or code changes are required.

Portability to Other Agents

Since skills are pure markdown, they work with other AI assistants. Copy skills to OpenCode, Gemini CLI, Cursor, or Kiro using:

for plugin in pm-*/; do
  mkdir -p .opencode/skills/
  cp -r "$plugin/skills/"* .opencode/skills/ 2>/dev/null
done

This extracts all 68 skills into a format consumable by any markdown-aware agent.

Summary

• PM Skills are markdown-based frameworks stored in the phuryn/pm-skills repository that provide structured product management methodologies to Claude.
• The architecture consists of 68 Skills (individual frameworks), 42 Commands (workflow orchestrators), and 9 Plugins (thematic bundles).
• Skills trigger automatically via phrase detection or explicitly via /plugin-name:skill-name syntax.
• Commands like /discover and /write-prd chain multiple skills into end-to-end workflows defined in commands/ markdown files.
• The system is fully extensible through markdown files alone—no code compilation required for new skills or commands.
• Skills are portable across AI assistants because they use standard markdown format.

Frequently Asked Questions

How do I install PM Skills in my Claude environment?

Install the marketplace via Claude Cowork, Claude Code CLI, or any compatible agent interface by referencing phuryn/pm-skills. The system pulls all nine plugins by default, though you can install individual plugins later if needed. Once installed, commands become available immediately and skills load automatically when relevant topics arise in conversation.

What is the difference between a skill and a command in PM Skills?

A skill is a single SKILL.md file containing one framework (like the PRD template in pm-execution/skills/create-prd/SKILL.md). A command is a workflow orchestrator that chains multiple skills together—such as /discover which runs ideation, assumption mapping, and prioritization sequentially. Skills are the building blocks; commands are the automated workflows.

Can I use PM Skills with AI assistants other than Claude?

Yes. Because skills are pure markdown files following a universal schema, they work with any AI assistant that can read markdown instructions. The repository includes a bash one-liner to copy all skills into .opencode/skills/ for use with OpenCode, Gemini CLI, Cursor, or Kiro. Simply copy the SKILL.md files into your preferred agent's skills directory.

How do I create a custom skill for my team's specific framework?

Create a new markdown file in the appropriate plugin's skills/ directory (or create a custom plugin folder). Follow the structure of existing files like pm-product-discovery/skills/brainstorm-ideas-existing/SKILL.md, including sections for trigger phrases, context, and output format. The skill becomes immediately available without restarting the system because Claude auto-detects new markdown files in skill directories.

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