Help us improve
Share bugs, ideas, or general feedback.
From stark
Maps user flows, information architecture, and state design (empty, loading, error, success) for products. Use for UX audits, onboarding, forms, dashboards, settings, or any usability improvement.
npx claudepluginhub f0d010c/stark --plugin starkHow this skill is triggered — by the user, by Claude, or both
Slash command
/stark:ux-designThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Use this skill when the product has to be understandable and usable, not just attractive. The output should make the user path clearer, reduce unnecessary decisions, and define states that survive real use.
Provides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Guides systematic root-cause debugging when tests fail, builds break, or unexpected errors occur. Provides a structured triage checklist to preserve evidence, localize, and fix issues instead of guessing.
Share bugs, ideas, or general feedback.
Use this skill when the product has to be understandable and usable, not just attractive. The output should make the user path clearer, reduce unnecessary decisions, and define states that survive real use.
Before designing screens, state the smallest useful context:
If any of these are unclear and materially affect the flow, ask one short question. If the answer is easy to infer from the request, infer it and continue.
Do not interrogate the user with a product-strategy questionnaire before helping. The skill should make useful assumptions, name them briefly, and move.
Write the minimum useful path:
Prefer fewer screens when the user is trying to finish one job. Prefer separate steps when the user is making risky, costly, or hard-to-reverse decisions.
Before visual design or code, write a compact brief. Keep it short enough to pass into another skill:
UX decision brief
- Job: ...
- User mode: ...
- Frequency/risk: ...
- Pattern: ...
- Primary action: ...
- Secondary actions: ...
- Core path: entry -> action -> feedback -> success
- Recovery path: ...
- Required states: empty, loading, partial, error, permission, success, long-running
- Handoff constraints: ...
This brief is the contract. The platform skill may change visual treatment, but it must not erase the chosen job, action hierarchy, state coverage, or recovery path.
Every production UI needs these states:
Do not ship only the happy path.
For public demos and generated proof projects, include at least one non-happy state in the visible UI: an empty state, blocked permission, failed sync, queued job, retry panel, stale data banner, or partial result. This makes the output feel like a real product instead of a polished poster.
Apply these rules:
For high-frequency tools, optimize scan speed, keyboard flow, saved filters, bulk actions, and stable layout. For consumer onboarding, optimize motivation, trust, and the shortest path to first value.
Use the product type to decide what "good UX" means:
| Product type | Optimize for | Avoid |
|---|---|---|
| SaaS dashboard | fast scanning, saved filters, drilldowns, clear priority, visible operational thesis | marketing-page spacing, decorative cards, hidden filters, generic CRM furniture |
| CRM/admin/internal tool | repeat speed, bulk actions, auditability, permissions, domain-specific task language | oversized empty space, playful copy, modal chains, interchangeable labels |
| Creative/editor tool | canvas focus, tool discoverability, undo/redo, stable panels | layout shifts, buried controls, destructive defaults |
| Marketplace/ecommerce | trust, comparison, price/shipping clarity, recovery | surprise costs, forced account creation, vague stock states |
| Onboarding/setup | first value, motivation, resumability, skip paths | long forms before value, fake progress, no return path |
| AI agent/tool run | plan preview, progress, artifacts, retry, stop/resume | invisible work, ambiguous completion, no trace of outputs |
If the request sounds like a real product people will use repeatedly, bias toward operational density and predictable navigation. If it is a one-off marketing page, bias toward clarity, brand memory, and conversion path.
If the product type clearly matches one of these contexts, read the matching brief before choosing the final UX pattern:
| Context | Read |
|---|---|
| Agent/tool run, background automation, long-running export/import | ../../references/ux-patterns/ai-agent-run.md |
| SaaS dashboard, CRM, admin panel, internal tool, support queue | ../../references/ux-patterns/operational-dashboard.md |
| First-run setup, trial activation, import/setup flow | ../../references/ux-patterns/activation-onboarding.md |
| Checkout, paywall, pricing, upgrade, plan comparison | ../../references/ux-patterns/checkout-upgrade.md |
| Editor, builder, canvas, creative tool, IDE-like surface | ../../references/ux-patterns/editor-canvas.md |
Use a brief only when the context fits. If no brief fits, proceed from the product type table and the user's actual constraints.
The brief should influence the UX decision brief, especially:
Do not copy a referenced app or blindly apply a pattern because it is common. A shipped screen is evidence that a real product team used a decision, not proof that it is best for every product.
Pick one pattern and name it before visual design:
| Need | Pattern |
|---|---|
| First run | Guided setup with skip/resume |
| Complex creation | Wizard with review step |
| Frequent operations | Command surface + saved views |
| Data-heavy work | Master/detail + filters + bulk actions |
| Monitoring | Dashboard with priority stack and drilldown |
| Settings | Searchable grouped settings + inline validation |
| Checkout/signup | Short form + transparent cost/risk + recovery |
| Collaboration | Activity timeline + comments + ownership |
| AI/tool execution | Plan preview + progress + artifacts + retry |
| Public demo/proof project | Product-specific job + proof surface + one non-happy state |
If two patterns fit, pick the one that reduces the riskiest failure mode. For example, choose a wizard over a single dense form when errors are costly, but choose command surface + saved views for repeated internal operations.
After the UX shape is clear, route to the platform skill:
../web-design/SKILL.md../windows-design/SKILL.md../apple-design/SKILL.md../android-design/SKILL.md../cross-platform-design/SKILL.mdPass the UX decisions into that skill as constraints. Do not let visual direction erase task flow, state coverage, or platform idioms.
When a contextual pattern brief was used, include its name in the handoff constraints so the platform skill knows which product behavior must survive visual design.
Before final delivery, check:
If any answer is no, fix the flow before polishing visuals.