How to Execute Commands from the phuryn/pm-skills Project

To execute commands from the phuryn/pm-skills project, install the marketplace via Claude Code CLI (claude plugin marketplace add phuryn/pm-skills) or Claude Cowork UI, then invoke any slash command (e.g., /review-resume) to trigger multi-step workflows defined in the repository's Markdown command files.

The phuryn/pm-skills repository is a marketplace of nine independent plugins that package product management skills (knowledge modules) and commands (executable workflows). When you execute commands from this project, Claude-based assistants interpret static Markdown files to run multi-step processes without server-side code execution.

Understanding the Command Architecture

Commands in phuryn/pm-skills follow a strict four-layer architecture defined in the repository's source files:

• Marketplace manifest: The root file .claude-plugin/marketplace.json lists all nine plugins and their versions, serving as the entry point for installation UIs.
• Plugin package: Each plugin resides under pm-{name}/ (e.g., pm-toolkit/) and contains a .claude-plugin/plugin.json manifest, a skills/ directory, and a commands/ directory.
• Skills: Data-driven "noun" modules stored as Markdown files (e.g., pm-toolkit/skills/review-resume/SKILL.md) that load automatically when Claude detects matching terminology.
• Commands: "Verb" files stored in commands/ directories (e.g., pm-toolkit/commands/review-resume.md) that chain skills into concrete workflows and are invoked with a leading slash (/).

The validate_plugins.py script checks manifest consistency, naming conventions, and version alignment across these layers.

Installing the phuryn/pm-skills Marketplace

You must install the marketplace before you can execute any commands. Choose the method matching your interface.

Via Claude Code CLI

Use the plugin CLI to register the marketplace and install specific plugins:

# Register the marketplace definition

claude plugin marketplace add phuryn/pm-skills

# Install individual plugins (example: pm-toolkit)

claude plugin install pm-toolkit@pm-skills

This copies the plugin's commands/ and skills/ directories into the assistant's runtime folder, making slash commands available immediately.

Via Claude Cowork UI

For GUI-based installation:

1. Open Customize → Browse plugins → Personal → +
2. Select Add marketplace from GitHub
3. Enter phuryn/pm-skills and confirm

All nine plugins become available in the interface once the marketplace JSON is parsed.

Executing Commands

After installation, invoke commands using the appropriate interface. The execution flow remains consistent: the assistant loads referenced skills from the skills/ directory, then processes the step-by-step prompts defined in the command's Markdown body.

Claude Code (Command Line)

In an active Claude Code session, type the slash command:

$ /review-resume

Claude will prompt for the resume file, auto-load the review-resume skill from pm-toolkit/skills/review-resume/SKILL.md, and return a structured review based on the workflow defined in pm-toolkit/commands/review-resume.md.

Codex-CLI (OpenAI)

Codex reads the same plugin.json manifests, allowing natural language invocation:

> Run a product discovery flow for "AI-powered meeting summarizer": brainstorm ideas → identify assumptions → prioritize → design experiments.

Codex internally locates the relevant skills (e.g., brainstorm-ideas, identify-assumptions) and produces output equivalent to the /discover command.

Validating Plugin Integrity

Before executing newly modified commands, run the repository validator to ensure manifest consistency:

python3 validate_plugins.py

This script checks that:

• All plugins listed in .claude-plugin/marketplace.json exist
• Command files reference valid skills
• Version strings align across manifests

If the script reports no errors, the marketplace is safe to use.

Summary

• Install first: Use claude plugin marketplace add phuryn/pm-skills (CLI) or the Cowork UI to register the marketplace.
• Invoke with slashes: Type commands like /review-resume or /write-prd to trigger workflows.
• Architecture: Commands live in commands/ directories, skills in skills/, and configuration in plugin.json files.
• Static execution: All logic resides in Markdown; no server-side code runs during command execution.
• Validate: Run python3 validate_plugins.py after any modifications to check file integrity.

Frequently Asked Questions

What is the difference between skills and commands in phuryn/pm-skills?

Skills are static knowledge modules stored in skills/{name}/SKILL.md that define domain expertise (like resume review criteria). Commands are executable workflows stored in commands/{name}.md that invoke one or more skills to produce artifacts. You execute commands; skills load automatically as dependencies.

How do I install only a specific plugin instead of the entire marketplace?

After adding the marketplace with claude plugin marketplace add phuryn/pm-skills, install individual plugins using claude plugin install {plugin-name}@pm-skills (e.g., claude plugin install pm-toolkit@pm-skills). This copies only that plugin's commands/ and skills/ directories without requiring the other eight.

Can I execute pm-skills commands without Claude Code?

Yes. The repository supports Claude Cowork (web UI) and Codex-CLI (OpenAI's tool). Claude Cowork uses the same slash commands, while Codex-CLI accepts natural language descriptions that map to the commands defined in the plugin.json manifests.

Where are command definitions stored in the repository?

Command definitions reside in each plugin's commands/ directory. For example, the resume review command is located at pm-toolkit/commands/review-resume.md, while its manifest entry exists in pm-toolkit/.claude-plugin/plugin.json. The root .claude-plugin/marketplace.json aggregates all available commands across the nine plugins.

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