pm-skills Project Structure: A Modular Marketplace Architecture

The pm-skills repository implements a plugin-based marketplace structure where nine domain-specific directories (like pm-product-discovery/ and pm-execution/) each contain self-contained skills defined in SKILL.md files and slash commands documented in commands/ folders, auto-validated by validate_plugins.py to ensure metadata consistency.

The phuryn/pm-skills project organizes product management expertise into discrete, auto-loaded modules that Claude and compatible AI assistants consume dynamically. Understanding this pm-skills project structure reveals how domain knowledge is encapsulated, validated, and exposed through standardized interfaces without requiring explicit imports or configuration.

Core Architecture: The Plugin Marketplace

The repository functions as a modular marketplace where each top-level directory represents a cohesive product management domain. The architecture distinguishes between two primary abstractions:

  • Skills – Self-contained knowledge blocks stored in skills/<skill-name>/SKILL.md files that define inputs, outputs, and execution logic
  • Commands – User-triggered workflows invoked with leading slashes (e.g., /discover, /write-prd) that chain one or more skills to produce end-to-end results

Plugins are auto-loaded by Claude when relevant capabilities are requested, eliminating manual configuration. The validate_plugins.py script enforces structural consistency across all modules, ensuring every skill has required metadata and matching command documentation where applicable.

Plugin Directory Breakdown

The repository contains nine specialized plugins following a consistent internal layout with skills/ and commands/ subdirectories.

pm-toolkit: General Utilities

The pm-toolkit/ directory houses cross-functional utilities including resume review and legal document drafting. Key files include:

pm-product-discovery: Ideation and Validation

The pm-product-discovery/ plugin manages the problem-space exploration phase with 13 skills and 5 commands:

  • Skills: brainstorm-ideas-new, identify-assumptions-new, prioritize-assumptions, brainstorm-experiments-new
  • Commands: /discover, /assumptions, /experiments

According to the source code, invoking /discover triggers an internal chain: brainstorm-ideas-new → identify-assumptions-new → prioritize-assumptions → brainstorm-experiments-new.

pm-product-strategy: Vision and Business Models

The pm-product-strategy/ directory contains 12 skills and 5 commands for strategic planning:

  • Skills: product-vision, pricing-strategy, business-model, competitive-analysis
  • Commands: /strategy, /pricing, /business-model

The /pricing command leverages the pricing-strategy skill to output competitive analysis, willingness-to-pay surveys, and price-elasticity calculations.

pm-execution: Day-to-Day Operations

The pm-execution/ plugin provides the largest capability set with 16 skills and 11 commands for operational product management:

  • Skills: create-prd, retro, okrs, roadmap, release-notes
  • Commands: /write-prd, /retro, /okrs, /roadmap

This plugin covers the full execution lifecycle from documentation through retrospectives.

Specialized Domain Plugins

Five additional plugins handle specific verticals:

  • pm-market-research/ – 7 skills (user-personas, market-sizing) and 3 commands (/research-users)
  • pm-data-analytics/ – 3 skills (sql-queries, cohort-analysis, ab-test-analysis) and 3 commands (/write-query)
  • pm-go-to-market/ – 6 skills (gtm-strategy, ideal-customer-profile) and 3 commands (/plan-launch)
  • pm-marketing-growth/ – 5 skills (north-star-metric) and 2 commands (/north-star)
  • pm-ai-shipping/ – 2 skills (shipping-artifacts, intended-vs-implemented) and 5 commands (/ship-check, /security-audit-static)

Repository-Wide Infrastructure Files

Beyond plugin directories, the root level contains critical governance and configuration files:

Validation and Testing

Agent Configuration

  • CLAUDE.md – The single source of truth for agent-specific guidance referenced by AGENTS.md
  • README.md – High-level project overview with installation instructions and a compact summary table of all plugins, skills, and commands

Practical Usage Examples

The following interactions demonstrate how the repository structure translates to end-user workflows:

Run a Full Discovery Workflow

/discover AI-powered meeting-summarizer for remote teams

This triggers the pm-product-discovery plugin's /discover command, executing the skill chain: brainstorm-ideas-new → identify-assumptions-new → prioritize-assumptions → brainstorm-experiments-new.

Generate a Pricing Strategy

/pricing Build a tiered pricing model for a B2B analytics platform

Invokes pm-product-strategy's /pricing command, utilizing the pricing-strategy skill for competitive analysis and price-elasticity calculations.

/draft-nda Draft a non-disclosure agreement for a cloud-provider collaboration covering IP clauses

Calls the pm-toolkit plugin's /draft-nda command, pulling the draft-nda skill templates.

Execute Security Audits

/security-audit-static src/api

Activates pm-ai-shipping's /security-audit-static command, combining the shipping-artifacts skill to map trust boundaries and the intended-vs-implemented skill to compare documented intent with actual code.

Summary

  • The pm-skills project structure organizes product management expertise into nine auto-loaded plugins: pm-toolkit, pm-product-discovery, pm-product-strategy, pm-execution, pm-market-research, pm-data-analytics, pm-go-to-market, pm-marketing-growth, and pm-ai-shipping
  • Each plugin contains skills/ directories with SKILL.md definition files and commands/ directories documenting slash-command interfaces
  • Skills are self-contained knowledge blocks; commands are user-facing workflows that chain multiple skills
  • validate_plugins.py enforces consistency across the marketplace, ensuring metadata completeness and proper command-skill relationships
  • CLAUDE.md serves as the canonical agent configuration, while tests/ provides automated validation coverage
  • Commands are invoked with leading slashes (e.g., /discover, /write-prd) and require no explicit imports or configuration

Frequently Asked Questions

What is the difference between a skill and a command in the pm-skills structure?

A skill is a self-contained knowledge block defined in a SKILL.md file (such as pm-toolkit/skills/review-resume/SKILL.md) that specifies inputs, outputs, and execution logic for a discrete task. A command is a user-facing interface documented in commands/*.md (such as pm-toolkit/commands/review-resume.md) that chains one or more skills together to produce an end-to-end result, invoked via slash syntax like /review-resume.

How does the pm-skills repository validate plugin consistency?

The repository uses validate_plugins.py to scan all nine plugin directories and enforce structural rules. The validator ensures every skill has required metadata (name, description), checks for matching command documentation where applicable, and verifies that SKILL.md files follow the universal schema. Unit tests in tests/test_validator.py provide automated coverage for this validation logic.

Where are skill definitions stored in the pm-skills project?

Skill definitions reside in skills/<skill-name>/ subdirectories within each plugin folder. For example, the resume review skill is located at pm-toolkit/skills/review-resume/SKILL.md, while product discovery skills live in pm-product-discovery/skills/. Each skill directory contains a SKILL.md file following a standardized schema that defines input arguments, response structures, and execution parameters.

How many plugins and commands are available in the pm-skills marketplace?

The repository contains nine plugins with a total of 64+ skills and 39+ commands. The largest plugin is pm-execution with 16 skills and 11 commands covering PRDs, OKRs, and retrospectives, while pm-ai-shipping is the most specialized with 2 skills and 5 commands focused on AI-generated code audits and static security analysis.

Have a question about this repo?

These articles cover the highlights, but your codebase questions are specific. Give your agent direct access to the source. Share this with your agent to get started:

Share the following with your agent to get started:
curl -s "https://instagit.com/install.md"

Works with
Claude Codex Cursor VS Code OpenClaw Any MCP Client

Maintain an open-source project? Get it listed too →