Guide · Agent skill
How to use the Prioritize Assumptions agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The Prioritize Assumptions skill (skill name: prioritize-assumptions) is a reusable set of instructions that teaches an AI agent to triage a list of product assumptions. It scores each assumption on impact and risk, places it in a 2x2 matrix, and recommends whether to defer, proceed, reject or test it. This guide explains what the skill does, how to install it, how to trigger it, and where its limits are, so you can use it on your own discovery work.
Prioritize Assumptionsphuryn/pm-skills · MIT Download the skill Quick start
- 1Download the skill. Download the ZIP from this page. It contains SKILL.md for prioritize-assumptions.
- 2Unzip into a skills folder. Unzip into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder for one project.
- 3Or upload in the Claude apps. In the Claude apps, open Settings, go to Capabilities, then Skills, and upload the skill.
- 4List your assumptions. Write each assumption as a single plain-language belief about your product or customers.
- 5Ask for a prioritization. Make a natural request such as triaging these assumptions or deciding what to test first, and attach any notes.
- 6Review the matrix. Check each assumption's category against your own judgement of impact, confidence and effort.
- 7Run the proposed experiments. For each assumption marked for testing, use the suggested experiment and its success metric.
What is an agent skill and what does this one do?
An agent skill is a folder containing a SKILL.md file. The file holds instructions that an agent such as Claude reads when a request matches the skill's description. This skill is part of the phuryn/pm-skills collection, under pm-product-discovery/skills/prioritize-assumptions, and is released under the MIT licence.
Given a list of assumptions about a product or idea, the agent evaluates each one on two dimensions. Impact combines the value of validating the assumption with the number of customers affected. Risk is defined in the skill as (1 - Confidence) × Effort. Each assumption then falls into one of four categories.
- Low Impact, Low Risk: defer testing until higher-priority assumptions are addressed
- High Impact, Low Risk: proceed to implementation
- Low Impact, High Risk: reject the idea as not worth the investment
- High Impact, High Risk: design an experiment to test it
How do you install the skill?
Download the ZIP file from the skill's page on this site. The package contains a single file, SKILL.md. You can install it in one of two ways.
For Claude Code, unzip the folder into your personal skills directory at ~/.claude/skills/, or into a project's .claude/skills/ directory if you want it limited to one project. For the Claude apps, open Settings, go to Capabilities, and upload the skill under Skills.
- Personal use across projects: ~/.claude/skills/
- Single project only: the .claude/skills/ folder inside that project
- Claude apps: Settings → Capabilities → Skills, then upload
How do you trigger the skill with a natural request?
You do not need to type a command. The skill's description tells the agent when to use it: triaging a list of assumptions, deciding what to test first, or applying the assumption prioritization canvas. Describe your situation in those terms and include the assumptions themselves.
The skill also says that if you provide files with assumptions or research data, the agent should read them first. So you can attach interview notes or a spreadsheet and ask it to prioritize what is in them.
- "Here are six assumptions about our onboarding idea. Which should we test first?"
- "Triage these assumptions using the assumption prioritization canvas."
- "Read the interview notes in this folder and tell me which assumptions are risky enough to test."
What output should you expect?
The skill asks the agent to present results as a prioritized matrix or table. For each assumption that needs testing, it proposes a small experiment that measures actual behaviour rather than opinions, with a clear success metric and threshold.
The skill draws on ICE and RICE scoring concepts. In ICE, Impact is Opportunity Score × number of customers, Confidence runs 1 to 10, and Ease runs 1 to 10. Opportunity Score is Importance × (1 − Satisfaction), normalised to 0 to 1, a formula attributed to Dan Olsen. RICE uses (Reach × Impact × Confidence) / Effort. The skill points to a separate prioritization-frameworks skill for full formulas and templates.
What do you need before you start?
The source lists no required tools, packages or accounts. You need an agent that can load skills and a list of assumptions written in plain language. The better your list, the more useful the result. Each assumption should name a specific belief, such as a customer segment wanting a feature or a price point being acceptable.
The skill does not run scripts and does not use the network, so it works only on what you give it.
What are the safety notes and limits?
The skill runs no scripts and makes no network calls, so it cannot pull live data. Its scores depend entirely on the judgement you supply for impact, confidence and effort. Because the output is a neat matrix, the numbers can look more precise than they are. Treat the ranking as a structured discussion aid, not a measurement.
The skill adds an analysis step. It does not build anything. If your roadmap is already validated and you need help with execution, this skill will not help much.
- No data source: scores reflect your judgement, so check them with the people who know the market
- Precision can be misleading: a matrix with numbers can look more rigorous than the inputs justify
- Not for execution planning: it does not produce code, tickets or delivery plans
What tips make the results better?
Give the agent context first: the product, the customer, and what you already know. Attach any research you have so the scores rest on evidence. Ask for the experiments explicitly, since the skill's experiment suggestions are the most actionable part of the output.
Review the categories before acting on them. A high-impact, high-risk assumption is the one most worth testing, so check whether the risk estimate is honest. Save the output as markdown when the analysis is long, as the skill suggests.
- Name each assumption as a single testable belief
- Supply your own confidence and effort estimates where you have them
- Ask for a success metric and threshold for every experiment
- Re-run the analysis after new evidence changes your confidence
What people use it for
- A product manager sorting twenty feature assumptions before a discovery sprint
- A founder deciding whether to validate pricing or the core workflow first
- A small team reviewing interview notes to find which beliefs are risky enough to test
- A student practising the assumption prioritization canvas on a case study
Frequently asked questions
Is the prioritize-assumptions skill free?
The skill is released under the MIT licence, as stated in its source listing. The source does not mention any price, so check the licence terms on the original repository before using it commercially.
What do I need to use this skill?
The source lists no requirements. You need an agent that can load skills and a list of assumptions. The skill runs no scripts and uses no network.
Where do I install the prioritize-assumptions skill?
Unzip the downloaded package into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. In the Claude apps, you can upload it under Settings, Capabilities, Skills.
How is this different from ICE or RICE?
ICE and RICE are scoring formulas for ranking items. This skill applies an Impact × Risk matrix to assumptions and adds a recommendation and experiment for each. It draws on ICE and RICE concepts, and it points to a separate prioritization-frameworks skill for full formulas.
Why are the scores not reliable on their own?
The skill has no data source, so impact, confidence and effort depend on the judgement you provide. The matrix can make those estimates look more precise than they are, so validate them with real evidence.
Can it analyse my research files?
Yes. The skill instructs the agent to read any files or research data you provide before prioritizing, so attach your notes or spreadsheets with your request.
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, founder or student who needs a structured way to decide which assumptions to test first and how. Skip it if you need hard numbers from data, or if your roadmap is already validated and you only need execution help.
Download the skill