How to Customize PM Skills Command Prompts and Parameters: A Complete Guide

You customize PM Skills commands by editing Markdown files in the pm-*/commands/ directory, where you can modify the argument-hint, description, and workflow steps without recompiling any code.

PM Skills is an open-source marketplace of plugins for product management workflows hosted at phuryn/pm-skills. Each plugin contains slash-style commands defined in plain Markdown, making customizing PM Skills command prompts and parameters as simple as editing text files. This architecture allows you to extend existing commands or create entirely new workflows by adjusting the argument definitions and skill invocations.

Understanding the PM Skills Architecture

The PM Skills repository organizes functionality into three distinct layers that separate metadata from implementation.

Marketplace Metadata Layer

The root /.claude-plugin/marketplace.json file lists all nine available plugins that users can install in a single step. This file acts as the central registry, identifying which plugin folders contain installable components.

Plugin Package Structure

Each plugin folder—such as pm-toolkit, pm-product-strategy, or pm-growth—follows a consistent structure. It contains a README.md describing the plugin's purpose, a .claude-plugin/plugin.json manifest that declares available skills and commands, and subdirectories for actual implementation files.

Command Definition Layer

Command files reside under pm-*/commands/ as plain Markdown documents. For example, pm-toolkit/commands/tailor-resume.md defines the /tailor-resume command, including the description, argument-hint, and step-by-step workflow that calls specific skills.

Editing Command Files to Customize Prompts

Customizing a command prompt requires modifying the associated Markdown file in your plugin's commands/ directory.

The pm-toolkit/commands/tailor-resume.md file contains the complete definition for the resume tailoring workflow. When you edit this file, you change how Claude-based assistants interact with users. The description field controls the help text, while the argument-hint specifies what data the assistant should request from the user.

Because these are pure Markdown files, you can version control your changes and deploy them immediately without rebuilding the plugin.

Adding New Parameters to Commands

Extending a command to accept additional parameters involves updating the argument-hint and modifying the workflow steps.

Follow this process to add a custom parameter:

1. Open the command file, such as pm-toolkit/commands/tailor-resume.md.
2. Locate the argument-hint line and add your new parameter.
3. Extend the workflow section to reference the new input.

For example, to add a tone parameter to the resume tailor command:

argument-hint: "<resume> + <job description> + <tone>"

After updating the hint, modify the workflow to pass the tone value to the underlying skill. Users can then invoke the enhanced command:

/tailor-resume "Resume text..." "Job description..." "formal"

The assistant automatically parses the new argument and incorporates it into the skill execution.

How Commands Execute Skills

Commands act as user-facing orchestrators that invoke reusable skills defined in separate Markdown files.

When a user types /tailor-resume, the assistant parses the argument-hint to determine required inputs. It then follows the workflow steps defined in the command file, which reference skills by name. For instance, the workflow might call the review-resume skill, which lives in pm-toolkit/skills/review-resume/SKILL.md.

The skill contains the actual implementation logic, while the command file handles the interaction pattern and parameter collection. This separation allows multiple commands to reuse the same skill with different prompts or parameter sets.

Installing and Testing Custom Commands

Before customizing commands, you must install the marketplace and relevant plugins using the Claude-Code CLI.

Install the marketplace and toolkit plugin:

# Install the marketplace and the toolkit plugin

claude plugin marketplace add phuryn/pm-skills
claude plugin install pm-toolkit@pm-skills

Run the resume-tailoring command with file uploads:

/tailor-resume [upload resume] Here’s the JD: [paste job description]

Or provide arguments directly:

/tailor-resume "My resume text…" "Job description text …"

Summary

• PM Skills uses a three-layer architecture: marketplace metadata, plugin packages, and command definitions.
• Command files are plain Markdown documents located in pm-*/commands/ directories.
• Customize prompts by editing the description, argument-hint, and workflow sections without code recompilation.
• Add parameters by updating the argument-hint and extending the workflow to pass new arguments to skills.
• Skills are reusable components defined in SKILL.md files that commands reference by name.

Frequently Asked Questions

Where are PM Skills command definitions stored?

Command definitions are stored as Markdown files in the commands/ subdirectory of each plugin folder, such as pm-toolkit/commands/tailor-resume.md. These files contain the argument-hint, workflow steps, and response templates.

Do I need to recompile code after customizing a command?

No. Because PM Skills defines commands in plain Markdown files, you can edit the argument-hint or workflow steps directly. Changes take effect immediately when the assistant reads the updated file, requiring no build or compilation step.

How do I add a new command to a PM Skills plugin?

Create a new Markdown file in the plugin's commands/ directory, define the argument-hint and workflow, and register it in the plugin's .claude-plugin/plugin.json manifest. The workflow should reference existing skills in the skills/ directory or define new ones.

Can multiple commands use the same skill?

Yes. Skills are reusable components defined in SKILL.md files within the skills/ directory. Multiple command files can reference the same skill by name, allowing you to create different interaction patterns or parameter sets that leverage identical underlying logic.

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