Lean UX
Collaborative, experiment-driven UX design
Practice collaborative, experiment-driven UX design using Jeff Gothelf and Josh Seiden's Lean UX methodology. This skill equips your AI agent with hypothesis-driven design, cross-functional collaboration techniques, and MVPs for validating UX decisions with real users.
npx skills add wondelai/skills/lean-ux --global 
Lean UX Jeff Gothelf & Josh Seiden View on Amazon
02 What it is
What is the Lean UX skill?
Lean UX moves design from deliverables to experiments, applying Jeff Gothelf and Josh Seiden’s book of the same name. The skill has your agent restate a feature request as a testable hypothesis — we believe this change produces that outcome, and we will know when this metric moves — then design the smallest thing that could disprove it.
03 Key concepts
Five principles your agent applies.
Drawn from Lean UX — not advice it repeats back to you, but decisions it makes while it works.
04 Fit
When to use it, and
when not to.
four symptoms it fixes, and four it doesn't.
Reach for it when
Your roadmap is a list of features with no statement of what any should change.
Design hands over comps, engineering builds them, and nobody checks whether the outcome arrived.
Stakeholders argue about the solution because the assumption underneath it was never written down.
You want to test a redesign before committing a quarter to building it.
Reach for something else when
You need interview technique that stops users being polite. The Mom Test is the sharper tool.
You want a weekly discovery rhythm and opportunity solution trees. Continuous Discovery goes deeper.
The business model is unproven, not just this screen. Lean Startup operates at that altitude.
You want a tested prototype by Friday. Design Sprint is the packaged five-day version.
05 Prompts
Say it like this.
Name the skill at the end of a prompt and the agent applies the framework rather than a vibe. Or say nothing — it loads on its own when a task matches.
Frame this feature idea as a Lean UX hypothesis with measurable outcomes using lean-ux skill
Plan a design studio session for our cross-functional team using lean-ux skill
Design a minimum viable experiment to test our onboarding redesign using lean-ux skill
Shift our team from output-based to outcome-based success metrics using lean-ux skill
06 About the author
07 Same stack
The stack it sits in.
Nine Design & UX skills. Most people run two: this one, and whichever fits the next problem.
All Design & UX skills08 Questions
Before you install.
How do I install the Lean UX skill?
Run npx skills add wondelai/skills/lean-ux --global. It takes about 30 seconds and needs no account. The Lean UX skill then works in Claude, Claude Code, Claude Cowork, Codex, Cursor, OpenClaw and Hermes Agent — anything that reads the open agentskills.io format — and your agent loads it on its own when a task calls for it. It is free and MIT-licensed, and the source is at https://github.com/wondelai/skills.
Which book is the Lean UX skill based on?
It packages Lean UX by Jeff Gothelf & Josh Seiden — lean UX pioneers, bestselling authors and product coaches. The skill distils the book's method into instructions your agent follows while it works, covering hypothesis-driven design, outcomes over outputs and cross-functional collaboration. It sits in the Design & UX part of the library.
How is it different from the Lean Startup skill?
Same loop, different altitude. Ries is testing whether the business should exist — the unit of learning is a pivot or a persevere, and the experiment might be a fake landing page for a product with no code behind it. Lean UX runs inside a team that already has a product, testing design decisions: this flow, this onboarding step, this navigation model. If you are questioning the company, use Lean Startup. If you are questioning the screen, use this.
Half the book is about cross-functional teams. Does it help a solo builder?
The collaboration chapters are largely moot when you are one person and an agent — there are no silos to break and design studio with a single participant is just sketching. What survives is most of the value: writing the assumption down before you build, framing it as a hypothesis with a named metric, and picking the smallest test that could kill it. Solo founders skip this discipline most, so it pays off disproportionately.
Do I need experiment infrastructure to run this?
No, and assuming you do is what stops teams starting. Most of the MVPs here are qualitative and cheap: a landing page that measures intent, a Wizard of Oz test where a human does the work the feature will eventually do, a concierge version for five customers. Statistical testing only earns its place once you have the traffic for it. Until then, five users watching a prototype answers the question faster.
Isn’t “outcomes over outputs” just a slogan?
It becomes one the moment you cannot name the metric. The skill is deliberately strict here — if the hypothesis ends in “users will be more engaged” with nothing instrumented behind it, you have written a wish, not an experiment. Being forced to specify what would change, by how much, and where you would see it is most of the value; a fair number of feature ideas do not survive that sentence.
09 Guides that use it
Put it to work.
Long-form playbooks where this skill does part of the job.
One skill, one command.
Install Lean UX, or the whole stack.
npx skills add wondelai/skills/lean-ux --global MIT licensed · Claude, Claude Code, Codex, Cursor, OpenClaw, Hermes · No account