How to Design Experiments to Test Assumptions Using the PM Skills Marketplace

The PM Skills Marketplace provides two specialized skills—brainstorm-experiments-existing and brainstorm-experiments-new—that guide you through clarifying assumptions, defining metrics, and outputting structured experiment plans with success thresholds.

The PM Skills Marketplace is a collection of Markdown-defined frameworks hosted in the phuryn/pm-skills repository. When you need to validate product hypotheses without wasting engineering resources, these deterministic skills orchestrate Claude-powered assistants through a rigorous experiment design process.

Understanding the Experiment Design Skills

The marketplace offers two complementary skills located in the pm-product-discovery plugin. Each skill follows a deterministic, step-by-step process defined in its respective SKILL.md file.

Brainstorm Experiments for Existing Products

The brainstorm-experiments-existing skill, defined in [pm-product-discovery/skills/brainstorm-experiments-existing/SKILL.md](https://github.com/phuryn/pm-skills/blob/main/pm-product-discovery/skills/brainstorm-experiments-existing/SKILL.md), designs low-effort experiments for live products. It guides you through idea clarification, assumption identification, and suggests concrete tests like first-click prototypes or fake-door experiments.

Brainstorm Experiments for New Products

The brainstorm-experiments-new skill, located at [pm-product-discovery/skills/brainstorm-experiments-new/SKILL.md](https://github.com/phuryn/pm-skills/blob/main/pm-product-discovery/skills/brainstorm-experiments-new/SKILL.md), focuses on lean-startup pretotype experiments. It forces you to create an XYZ hypothesis (e.g., "X% of Y will do Z") and suggests validation methods like landing pages, explainer videos, or pre-order tests that require "skin-in-the-game" from potential customers.

The Experiment Design Workflow

Both skills accept an argument string ($ARGUMENTS) describing your product context, assumptions, and file attachments. The execution follows a strict five-step process:

1. Read user-provided files (PRDs, analytics CSVs, or data dumps) automatically loaded by the assistant.
2. Clarify the idea (existing) or create an XYZ hypothesis (new) to frame the core assumption.
3. Suggest 2-3 concrete experiments targeting behavioral validation rather than opinion surveys.
4. Specify for each experiment:
   • Assumption being tested
   • Experiment description
   • Metric to capture (behavioral, not attitudinal)
   • Success threshold (e.g., "≥ 15% sign-ups")
5. Output a markdown table ready for direct insertion into your PRD or team wiki.

Running Experiments Through Commands and Skills

You can invoke these skills through multiple interfaces depending on your environment.

Using Claude Code CLI

Install the discovery plugin once, then invoke the skill via natural language:

# Install the plugin

claude plugin install pm-product-discovery@pm-skills

# Request experiment design for an existing product

claude chat "Design experiments for the following assumptions:
- Reducing onboarding churn by improving the welcome flow
- Increasing daily active users by adding a social share button"

The assistant automatically loads brainstorm-experiments-existing and returns a structured markdown table with metrics and success thresholds.

Using Slash Commands in Claude Cowork

In Claude Cowork environments, use explicit slash commands to trigger the skill:

/brainstorm experiments existing — We need to reduce churn in our onboarding flow

This command triggers the same skill, reads any attached files (e.g., an onboarding analytics CSV), and produces a structured experiment plan.

Chaining the Full Discovery Flow

The /discover command, defined in [pm-product-discovery/commands/discover.md](https://github.com/phuryn/pm-skills/blob/main/pm-product-discovery/commands/discover.md), orchestrates a complete discovery workflow by chaining multiple skills:

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

Invoke it with a single product concept:

/discover AI-powered meeting summarizer for remote teams

The final output includes both a prioritized assumption matrix and a set of lean-startup experiments targeting the highest-risk assumptions.

Direct Integration with OpenCode

For programmatic access, call the skill directly from Python scripts:

from opencode import Claude

assistant = Claude()
response = assistant.ask("""
Design experiments for the following assumptions (existing product):
1. Users abandon the checkout after entering payment details.
2. Adding a progress bar will increase completion rates.
""")
print(response)   # Markdown table with experiments, metrics, thresholds

Prioritizing Assumptions Before Experimentation

Before designing experiments, use the prioritize-assumptions skill located at [pm-product-discovery/skills/prioritize-assumptions/SKILL.md](https://github.com/phuryn/pm-skills/blob/main/pm-product-discovery/skills/prioritize-assumptions/SKILL.md). This framework scores assumptions on an Impact × Risk matrix and automatically tags high-impact/high-risk items that require immediate experimentation.

Running this skill first ensures you only design experiments for assumptions that actually matter, avoiding wasted effort on low-risk optimizations.

Summary

• The PM Skills Marketplace provides brainstorm-experiments-existing and brainstorm-experiments-new skills to validate assumptions for live and hypothetical products.
• Each skill accepts an argument string ($ARGUMENTS) and outputs a markdown table containing assumptions, metrics, and success thresholds.
• The /discover command chains the full workflow from ideation to experiment design.
• Skills are plain Markdown files that work across Claude Code, Claude Cowork, OpenCode, and Gemini CLI.
• Always run prioritize-assumptions first to target high-risk items.

Frequently Asked Questions

What is the PM Skills Marketplace?

The PM Skills Marketplace is a repository of Markdown-defined skills and commands hosted at phuryn/pm-skills. It provides self-contained frameworks that Claude-powered assistants can invoke to perform structured product management tasks, including experiment design, assumption prioritization, and PRD generation.

How do I choose between existing and new product experiment skills?

Use brainstorm-experiments-existing when your product is already in the market and you need to optimize specific metrics like churn or conversion. Use brainstorm-experiments-new when validating a pre-launch concept or pivot, as it forces XYZ hypothesis creation and lean-startup pretotypes like landing pages or pre-orders.

Can I use these skills outside of Claude?

Yes. Because the skills are plain Markdown files, they function in any assistant environment that can read SKILL.md files and render the output, including OpenCode, Gemini CLI, and custom implementations. The only requirement is that the underlying LLM can process the structured instructions contained in the files.

What makes a good experiment metric in this framework?

The framework requires behavioral metrics over attitudinal ones. Instead of asking "Would you use this?" (opinion), the skills force you to measure actions like sign-up rates, button clicks, or pre-order completions. Each experiment must define a specific success threshold (e.g., "≥ 15% click-through rate") before execution begins.

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