product-discovery

Overview of product-discovery skill

What product-discovery does

The product-discovery skill helps an AI agent run structured product discovery before a team commits engineering time. It is built for validating opportunities, mapping risky assumptions, planning discovery sprints, and deciding whether to proceed, pivot, or stop. Instead of asking for “product ideas,” the skill pushes the work toward evidence: outcomes, opportunities, assumptions, experiments, and learning decisions.

Best fit for Product Management work

Use product-discovery for Product Management when you need a repeatable discovery workflow around unclear customer problems, new feature bets, market-facing experiments, or early solution concepts. It is especially useful for product managers, founders, product designers, UX researchers, and cross-functional squads who want an AI assistant to structure discovery artifacts rather than generate a generic roadmap.

What makes this skill different

The skill centers on practical discovery frameworks: Opportunity Solution Trees, assumption mapping, Jobs-to-be-Done, Kano, design sprint thinking, and experiment planning. Its strongest differentiator is the included scripts/assumption_mapper.py, which can prioritize assumptions from a CSV using risk and certainty scores and suggest suitable validation tests by assumption category.

When it is not the right tool

Do not install this skill if you mainly need delivery planning, sprint backlog grooming, PRD formatting, growth copy, or analytics instrumentation. product-discovery is most valuable before delivery, when the team still needs to clarify the user problem, identify the riskiest beliefs, and choose low-cost validation methods.

How to Use product-discovery skill

product-discovery install and repository path

Install the skill from the GitHub repository path used by your skill manager. A typical install command is:

npx skills add alirezarezvani/claude-skills --skill product-discovery

The source lives at:

product-team/skills/product-discovery

After install, read SKILL.md first, then references/discovery-frameworks.md, and finally scripts/assumption_mapper.py if you plan to score assumptions from a CSV. The repository has no separate README.md or metadata.json in this skill folder, so the main operating instructions are concentrated in those three files.

Inputs the skill needs

For strong product-discovery usage, give the agent more than a feature request. Include:

• Target outcome: the metric or behavior you want to improve
• User segment: who has the problem and in what context
• Current evidence: interviews, support tickets, analytics, sales notes, churn reasons
• Candidate opportunity: the pain, need, or job to be done
• Constraints: timeline, team capacity, compliance, market, technical limits
• Decision needed: proceed, pivot, stop, run interviews, prototype, or design an experiment

Weak prompt:

Help us validate a new onboarding feature.

Stronger prompt:

Use product-discovery to plan discovery for reducing activation drop-off from 42% to 30% in 8 weeks. Segment: self-serve B2B admins setting up their first workspace. Evidence: 12 support tickets mention confusing permissions; analytics show most drop-offs happen before inviting teammates. We are considering an onboarding checklist but are unsure if the real opportunity is permissions clarity, team invitation anxiety, or lack of perceived value. Produce an Opportunity Solution Tree, risky assumptions, and a 1-week validation plan.

Suggested workflow for first use

Start by asking the skill to define one measurable outcome and build an Opportunity Solution Tree: outcome → opportunities → solution ideas → experiments. Then ask it to separate evidence-backed opportunities from internal opinions. Next, generate desirability, viability, feasibility, and usability assumptions. Finally, convert the highest-risk assumptions into interviews, prototype tests, fake-door tests, pricing tests, or technical spikes.

If you have assumptions ready, create a CSV with these columns:

assumption,category,risk,certainty

Use values from 0 to 1 for risk and certainty, then run:

python3 scripts/assumption_mapper.py assumptions.csv

The script prioritizes high-risk, low-certainty assumptions and suggests a validation test type.

Practical prompt patterns

Ask for decision-ready outputs, not just frameworks. Good requests include:

• “Create an OST and mark which branches need more evidence.”
• “Turn these interview notes into opportunity themes and confidence levels.”
• “Map assumptions and identify the cheapest test for each top risk.”
• “Design a 1-week discovery sprint with daily evidence reviews.”
• “Define stop, pivot, and proceed criteria before we run tests.”

The skill performs better when you force explicit evidence labels: observed behavior, direct quote, metric, internal opinion, or unknown.

product-discovery skill FAQ

Is product-discovery only for product managers?

No. The product-discovery skill is framed for Product Management, but it is useful for founders, designers, researchers, growth teams, and technical leads involved in de-risking product bets. The key requirement is that the user can provide context about customers, business goals, and constraints.

How is it better than an ordinary discovery prompt?

A normal prompt may produce a list of questions or experiments. product-discovery gives the agent a more specific operating model: measurable outcome, Opportunity Solution Tree, assumption categories, risk/certainty scoring, problem validation, solution validation, and discovery sprint decisions. That structure reduces guesswork and makes outputs easier to compare across opportunities.

Does it replace user research?

No. It helps plan and synthesize discovery, but it cannot replace interviews, behavioral data, prototype testing, or market evidence. Treat its outputs as hypotheses and operating plans. The quality of the result depends heavily on the evidence you provide and the team’s willingness to reject weak opportunities.

What should beginners read first?

Start with SKILL.md to understand the workflow, then read references/discovery-frameworks.md for the framework definitions. If you are new to discovery, focus on three concepts first: Opportunity Solution Tree, Jobs-to-be-Done interview framing, and the assumption prioritization matrix. Use the Python script only after you understand what each assumption means.

How to Improve product-discovery skill

Improve product-discovery results with better evidence

The fastest way to improve product-discovery output is to provide raw evidence, not polished conclusions. Include interview excerpts, behavioral metrics, support tickets, lost-deal reasons, usage funnels, or prototype observations. Ask the agent to distinguish “evidence” from “interpretation” so your team does not accidentally validate an internal preference.

Avoid common failure modes

Common weak outputs include solution-first trees, vague assumptions, oversized experiments, and interview plans that ask leading questions. Counter these by prompting for:

• Opportunities before solutions
• Assumptions stated as testable claims
• Smallest credible experiment
• Success and failure thresholds
• What decision will change after the test

For example, replace “users want better onboarding” with “new workspace admins fail to invite teammates because they do not understand permission consequences.”

Iterate after the first output

Do not treat the first result as final. Ask the skill to critique its own discovery plan against cost, speed, evidence quality, and decision impact. Then ask it to remove experiments that do not change a real decision. A useful second-pass prompt is:

Review this discovery plan. Identify assumptions that are too vague, experiments that are too expensive, and places where we are testing preference instead of behavior. Revise into a 5-day plan with clear proceed, pivot, and stop criteria.

Customize the skill for your team

For better long-term use, add team-specific examples: your product metrics, customer segments, research templates, experiment standards, and decision thresholds. If your organization has strict compliance, enterprise sales cycles, marketplace dynamics, or hardware constraints, include those in prompts. product-discovery is strongest when its general frameworks are grounded in your actual operating environment.