Guide · Agent skill
How to use the Brainstorm Product Experiments agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The brainstorm-experiments-existing skill is a set of instructions that teaches an AI agent to help a product team design low-effort experiments for an existing product. It is useful when you have a feature idea and want to test its assumptions cheaply before committing engineering time. By the end of this guide you will know what the skill does, how to install it, how to trigger it with a plain request, and what to check in its output.
Brainstorm Product Experimentsphuryn/pm-skills · MIT Download the skill Quick start
- 1Download the ZIP. Get the brainstorm-experiments-existing ZIP file from this page.
- 2Unzip the skill folder. Unzip the file so that the folder containing SKILL.md is in place.
- 3Place it in a skills folder. Move the folder to ~/.claude/skills/ for personal use or to .claude/skills/ inside a project.
- 4Or upload it in the Claude apps. In the Claude apps, open Settings, go to Capabilities, then Skills, and upload the skill.
- 5Describe your idea. Write a request that names the feature and the assumptions you want to validate.
- 6Review the experiment plan. Check each assumption's experiment, metric and success threshold before you run anything.
What is an agent skill and what does this one do?
An agent skill is a folder containing a SKILL.md file with instructions that an AI agent can load when a task matches the skill's description. This skill is part of the pm-product-discovery set from phuryn/pm-skills and is released under the MIT licence.
Once the skill is active, the agent works through your idea step by step. It clarifies what you want to build and which assumptions need validating, then suggests experiments for each assumption. Methods it considers include first-click testing or task completion with a prototype, feature stubs or fake door tests, technical spikes, A/B tests on production with risk mitigation, Wizard of Oz approaches, and survey-based validation focused on behaviour rather than opinion.
- For each experiment it specifies the assumption, the experiment, the metric, and the success threshold.
- It emphasises measuring actual behaviour rather than users' opinions.
- For production tests such as A/B tests, it explains risk mitigation strategies.
- If you provide files such as PRDs, assumption lists or designs, the agent reads them first.
Install the skill from the ZIP file
Download the ZIP file for this skill from this page and unzip it. The skill is a folder that contains SKILL.md. Place that folder in one of these locations:
- Personal skills folder: ~/.claude/skills/ so the skill is available across your projects.
- Project skills folder: .claude/skills/ inside a specific project, so the skill is available only to that project.
- Claude apps: open Settings, go to Capabilities, then Skills, and upload the skill there.
Trigger the skill with a natural request
You do not need a special command. Describe your idea and say that you want to validate assumptions or plan experiments. The skill's description mentions validating assumptions, testing feature ideas cheaply, and planning product experiments, so requests using those ideas are the most likely to match.
The skill's instructions refer to $ARGUMENTS, which holds the subject you describe. Give the agent enough context to work with: the feature, who it is for, and what you believe must be true for it to succeed.
- "Help me design experiments to test whether users want a bulk-export feature in our existing dashboard."
- "We think customers will pay for an annual plan. What cheap experiments could validate that assumption?"
- "Plan product experiments for our new onboarding checklist. Here is our assumption list." (attach the file)
What to expect in the output
The source suggests presenting experiments in a clear table or structured format, and saving the result as markdown if it is substantial. Each assumption should come with its experiment, metric and success threshold, so you can review them as a plan.
Treat the success thresholds as proposals. The skill states that the expected value applies if you are right, so you and your team need to decide whether those numbers fit your product and business.
What it needs and safety notes
The skill lists no required tools or packages. It does not run scripts and does not use the network. It consists of a single SKILL.md file with instructions only.
Because it does not run the tests, you still need to carry out the experiments yourself. The skill also stresses responsible testing: do not put users or the business at risk, and plan risk mitigation before running any A/B test on live users.
- Requires: nothing listed.
- Runs scripts: no.
- Uses network: no.
- It only produces a plan. It does not run A/B tests or collect metrics.
Tips for better experiment plans
The quality of the plan depends on the quality of your input. List your assumptions explicitly, separate the riskiest ones, and say which ones you can test without building anything.
- Give the agent a short list of assumptions rather than only a feature name.
- Ask for behavioural metrics and say which ones your analytics already track.
- Share constraints such as traffic volume, engineering time or whether you have live users to test with.
- Ask the agent to explain the risk mitigation for any production test.
- Have an experienced experiment designer review the final plan before you act on it.
What people use it for
- Planning a fake door test to see whether users click on a feature that does not yet exist.
- Choosing between a technical spike and a prototype test before estimating engineering work.
- Designing an A/B test on production with risk mitigation steps for a new existing-product feature.
- Structuring a Wizard of Oz test where a person simulates a feature manually to observe behaviour.
- Turning a list of assumptions into a table of experiments, metrics and success thresholds for a team review.
Frequently asked questions
How do I install the brainstorm-experiments-existing skill?
Download the ZIP file from this page and unzip it. Place the folder containing SKILL.md in ~/.claude/skills/ for personal use, or in .claude/skills/ inside a project. In the Claude apps, you can instead upload it under Settings, Capabilities, then Skills.
Is this agent skill free to use?
The skill is released under the MIT licence, as stated on this page. Check the licence file in the source repository for the full terms.
What does the skill require to run?
The skill lists no requirements. It runs no scripts and does not use the network, and it consists of a SKILL.md file with instructions.
Does the skill run A/B tests or collect metrics for me?
No. The skill only produces a plan. It suggests experiments and defines the metric and success threshold for each, but you run the tests and gather the data yourself.
How is this different from just asking an AI for experiment ideas?
The skill gives the agent a fixed working method: clarify the idea, suggest experiments for each assumption, apply principles such as measuring behaviour and managing risk, and specify assumption, experiment, metric and threshold for each one. That structure is the main difference, though the output is still a generic methodology suggestion.
Why is the skill not showing up or not triggering?
Check that the folder containing SKILL.md sits directly in ~/.claude/skills/ or .claude/skills/, or that the uploaded skill appears under Settings, Capabilities, then Skills. Then phrase your request around validating assumptions, testing feature ideas cheaply, or planning product experiments, which match the skill description.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you are a product manager, product team or founder who wants a structured, low-effort experiment plan before committing engineering time. Pick something else if you need someone to run tests, analyse real data or guarantee statistical rigour, and have the plan reviewed by an experienced experiment designer either way.
Download the skill