What Distinguishes Skills from Commands in the PM Skills Architecture

Skills are modular knowledge blocks that encode product management frameworks, while commands are user-triggered workflows that orchestrate multiple skills into complete end-to-end processes.

The phuryn/pm-skills repository implements a modular AI architecture that separates domain knowledge from workflow orchestration. Understanding what distinguishes skills from commands in the PM Skills architecture is essential for effectively leveraging this open-source product management toolkit.

Core Architectural Concepts

What Are Skills

Skills are standalone, reusable modules that encapsulate specific product management knowledge, frameworks, or analytical capabilities. According to the phuryn/pm-skills source code, each skill resides as a Markdown file under pm-*/skills/…/SKILL.md and defines a single domain concept—such as brainstorm-ideas-new or product-vision.

Skills auto-load when the model detects relevant domain context, or users can force-load them explicitly using slash-prefixed syntax (e.g., /skill-name or /plugin-name:skill-name). This design makes skills language-agnostic and portable across different AI models including Claude, Gemini, and OpenCode.

What Are Commands

Commands are end-to-end workflows triggered explicitly by users via slash-prefixed invocations (e.g., /discover, /strategy). Defined in pm-*/commands/…md files, commands function as scripts that chain multiple skills together to accomplish complex product management tasks.

When executed, a command invokes its constituent skills in sequence, passing the output of one skill as input to the next. This compositional approach allows a single command to execute multi-step processes—such as running a full discovery cycle—while reusing existing skill definitions.

Key Differences Between Skills and Commands

The PM Skills architecture maintains a strict separation between knowledge storage and workflow orchestration:

• Definition Location: Skills live in pm-*/skills/…/SKILL.md files (e.g., pm-product-discovery/skills/brainstorm-ideas-new/SKILL.md), while commands reside in pm-*/commands/…md files (e.g., pm-product-discovery/commands/discover.md).

• Loading Mechanism: Skills load automatically when domain-relevant context is detected, or via explicit /skill-name invocation. Commands require explicit user triggers using /command-name syntax and do not auto-load.

• Composition Model: Skills are atomic and standalone—encapsulating a single framework like the product-vision skill. Commands are composite—chaining multiple skills (e.g., /discover calls brainstorm-ideas-new → identify-assumptions-new → prioritize-assumptions → brainstorm-experiments-new).

• Portability: Skills work across different AI platforms using natural language instructions. Commands use Claude-specific slash syntax and require translation for other agents.

How Skills and Commands Interact

The interaction between skills and commands follows a pipeline pattern. When a user invokes a command, the system:

1. Loads the command definition from the corresponding Markdown file
2. Identifies the ordered list of skills specified in the command
3. Executes each skill sequentially, maintaining context between steps
4. Returns the final aggregated output to the user

After execution, commands suggest the next logical command in the workflow, creating guided product management journeys while keeping the underlying skills decoupled and reusable.

File Structure and Implementation

The repository organizes these components through a predictable directory structure:

• Skill definitions: pm-product-discovery/skills/brainstorm-ideas-new/SKILL.md encodes a single brainstorming framework
• Command definitions: pm-product-discovery/commands/discover.md orchestrates multiple skills into the /discover workflow
• Cross-domain reuse: pm-toolkit/skills/review-resume/SKILL.md demonstrates skills that serve multiple command contexts

Practical Usage Examples

Loading a Single Skill Directly

To access a specific framework without running a full workflow, force-load the skill using its slash-prefixed name:

/product-vision

This invokes the product-vision skill defined in pm-product-strategy/skills/product-vision/SKILL.md, loading only that specific framework.

Executing a Multi-Step Command

To run a complete discovery process that chains multiple skills:

/discover AI-powered meeting summarizer for remote teams

This executes the /discover command defined in pm-product-discovery/commands/discover.md, which internally calls:

1. brainstorm-ideas-new
2. identify-assumptions-new
3. prioritize-assumptions
4. brainstorm-experiments-new

Each skill receives the output from the previous step, creating a cohesive discovery output while maintaining individual skill modularity.

Skill Definition Example

The brainstorm-ideas-new skill in pm-product-discovery/skills/brainstorm-ideas-new/SKILL.md contains:

# Brainstorm Ideas (New Products)

You are an expert product strategist.  
Generate 5 distinct ideas for a brand‑new SaaS product that solves a common pain point.

This standalone module can be called by any command requiring ideation capabilities.

Summary

• Skills are atomic knowledge modules stored in SKILL.md files that auto-load based on context and encode specific product management frameworks
• Commands are executable workflows stored in commands/*.md files that chain multiple skills using slash-prefixed triggers like /discover
• The architecture separates knowledge (skills) from process (commands), enabling skills to be reused across multiple commands without modification
• Skills are language-agnostic and portable across AI models, while commands use Claude-specific syntax and require translation for other platforms
• Commands execute skills sequentially, passing context between steps to create end-to-end workflows while suggesting next logical actions

Frequently Asked Questions

What file format defines a skill in PM Skills?

Skills are defined as Markdown files named SKILL.md located in directories under pm-*/skills/. For example, the brainstorming skill resides at pm-product-discovery/skills/brainstorm-ideas-new/SKILL.md. These files contain natural language instructions that encode specific product management frameworks.

How do I invoke a command versus loading a skill directly?

Commands require explicit slash-prefixed invocation like /discover or /strategy, triggering multi-step workflows defined in command files. Skills can be force-loaded using /skill-name syntax (e.g., /product-vision) or load automatically when the model detects relevant domain context in the conversation.

Can I create a custom command that uses existing skills?

Yes. Create a new Markdown file in pm-*/commands/ and define the workflow by listing existing skills in execution order. The /discover command demonstrates this pattern by chaining brainstorm-ideas-new, identify-assumptions-new, prioritize-assumptions, and brainstorm-experiments-new without modifying the underlying skill definitions.

Are skills and commands interchangeable between different AI models?

Skills are portable across AI platforms because they use natural language instructions. Commands, however, rely on Claude-specific slash-command syntax. While other agents like Gemini or OpenCode can read command files, they must translate the workflow definitions into their own execution language to run them.

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