UX Research & Usability Testing Skills for Claude Code (2026)

Claude Code can generate a UI in seconds, but it can't tell you whether real users understand it - that gap is what UX research and usability-testing skills close. How to design a UI with Claude Code covers the visual stack; this guide covers the research layer underneath it.

Key takeaways

• ux-designer by shubhamsaboo (3,560 installs, 113,803 stars - skills.sh registry) embeds UX research, persona creation, accessibility and inclusive-design rules into every Claude Code session.
• usability-testing by refoundai (1,683 installs, 1,030 stars) runs structured usability tests that surface friction before a feature ships in full.
• user-personas by phuryn (1,575 installs, 12,332 stars) turns raw survey data and interview notes into three refined, actionable personas.
• customer-journey-map - two independent builds: deanpeters at 1,565 installs/4,957 stars, and phuryn's version at 1,551 installs/12,332 stars - maps how users move from first awareness to adoption.
• Pair the research layer with design system skills and accessibility skills so findings turn into shipped fixes, not just reports.

Why do UI-generation skills need a research layer?

A skill that generates polished components still guesses at what users actually need - it has no signal on where they get confused, abandon a flow, or misread a label. Research skills give Claude Code (and the team reviewing its output) that signal before code ships, not after a support ticket arrives. The catalog carries 20+ listings matching "UX research" alone (skills.sh registry, June 2026), confirming builders are actively packaging this workflow into reusable skills rather than doing it ad hoc.

What is the best UX research skill for Claude Code?

ux-designer by shubhamsaboo leads with 3,560 installs and 113,803 GitHub stars (skills.sh registry) - the highest-starred repo in this niche. It bundles UX research prompts, persona generation, accessibility checks and inclusive-design rules into one skill, so a single install covers most of the research-to-design handoff.

claude skills install ux-designer

Two close alternatives specialize instead of bundling: ux-researcher-designer by davila7 (1,004 installs, 27,831 stars) focuses on generating data-driven personas that "prevent building for imagined users," and ux-researcher-designer by alirezarezvani (822 installs, 17,479 stars) adds journey maps and usability-test-plan generation on top. See the full design system pillar for how these fit into a larger stack.

How do you actually run a usability test with Claude Code?

Install a dedicated test-runner rather than relying on a general design skill for this step. usability-testing by refoundai (1,683 installs, 1,030 stars) runs structured tests designed to reveal friction "before investing in full development," and replay-ux-research by getsentry (438 installs, 803 stars) takes the opposite angle - it analyzes real Sentry session replays to understand actual user behavior rather than simulated tests. Use the replay-based skill once a feature is live; use the structured tester before it ships.

How do personas and journey maps fit into an AI coding workflow?

They give Claude Code a concrete subject to design for instead of an abstract "the user." user-personas by phuryn (1,575 installs, 12,332 stars) converts survey data or interview notes into three refined personas in one pass, and customer-journey-map (built independently by both deanpeters at 1,565 installs and phuryn at 1,551 installs) visualizes the friction points along the way. Feed the persona and journey output back into a prompt ("design this settings page for the persona and journey map above") to keep every generated component grounded in real research instead of guesswork.

A complete starter stack

• ux-designer (3,560 installs) - the all-in-one research + persona + accessibility layer
• user-personas (1,575 installs) - dedicated persona generation from raw research data
• customer-journey-map (1,565 installs) - visualizes the end-to-end experience
• usability-testing (1,683 installs) - structured pre-ship testing
• replay-ux-research (438 installs) - post-ship behavior analysis from real session replays

Common pitfalls

• Skipping personas and going straight to UI generation - Claude Code will default to designing for an imagined power user, not your actual audience.
• Running usability tests only after launch - usability-testing is built to run before full development investment, not as a post-mortem.
• Treating research skills as a one-time setup instead of an input you re-run each time requirements shift.

Browse the full research and design cluster in the UX design hub or the accessibility hub.

Common questions