arrange
by pbakausThe arrange skill helps improve UI layout, spacing, grouping, and visual hierarchy. Learn when to use it, how arrange install works from the GitHub repo, and how to apply arrange usage with /frontend-design context for better UI Design decisions.
This skill scores 68/100, which means it is listable for directory users but with clear caveats: the trigger intent is strong and the design critique focus is real, yet execution still depends heavily on another skill and on the agent translating qualitative guidance into concrete changes.
- Strong triggerability from a specific description covering layout, spacing, hierarchy, crowding, alignment, and composition issues
- Provides substantive evaluation criteria around spacing, visual hierarchy, and grid structure rather than placeholder advice
- Includes explicit prerequisite routing to /frontend-design and /teach-impeccable, which helps situate the skill in a broader workflow
- Operational clarity is limited because the skill requires another skill's context-gathering protocol and gives no local quick-start or concrete execution artifacts
- No scripts, examples, code fences, or referenced files are provided, so agents must infer how to turn critique into specific implementation changes
Overview of arrange skill
What arrange does
The arrange skill is a UI layout critique-and-improvement skill for fixing spacing, hierarchy, grouping, and overall visual rhythm. It is best used when a screen feels crowded, flat, repetitive, or “off” even if the UI technically works.
Who should use arrange
arrange is most useful for designers, frontend developers, and AI-assisted builders who already have a screen, component, or rough mockup and want better composition without jumping straight into a full visual redesign. It is especially relevant for arrange for UI Design workflows where the main problem is structure, not branding.
The real job-to-be-done
Most users do not need abstract design theory. They need the AI to look at a weak layout, diagnose why it feels monotonous or unclear, and propose concrete improvements to spacing, grouping, and hierarchy. The arrange skill is built for that narrower job.
What makes this skill different
Unlike a generic “make this prettier” prompt, arrange focuses on spatial decisions: consistent-but-not-flat spacing, stronger grouping, clearer scan paths, and less mechanical grid repetition. It also has an important dependency: it expects design context from /frontend-design, and if that context does not exist yet, it requires /teach-impeccable first.
Best-fit and misfit cases
Use arrange when:
- cards all look the same
- spacing feels arbitrary or uniformly padded
- hierarchy is weak
- sections blend together
- alignment is technically correct but visually dull
Do not choose arrange first when the problem is mainly:
- color palette
- typography system
- brand direction
- accessibility implementation details
- missing product strategy or unclear content
How to Use arrange skill
Install context for arrange
The repository does not publish its own install command inside SKILL.md, so directory users typically add it through their skill manager using the GitHub source path for the arrange skill. If your environment supports GitHub skill installs, use the repo URL and map the skill name to arrange.
A common pattern is:
npx skills add https://github.com/pbakaus/impeccable --skill arrange
If your setup uses a different installer, the important part is the source repository: pbakaus/impeccable, skill path .claude/skills/arrange.
Read this before first use
For arrange install decisions, the key thing is that this skill is not fully standalone. The skill explicitly says to invoke /frontend-design first because that parent skill contains design principles and a context-gathering protocol. If no design context exists yet, it says you must run /teach-impeccable before continuing.
That means adoption is easiest if you are already using the broader impeccable skill set rather than cherry-picking one isolated prompt.
What input arrange needs
The strongest arrange usage starts with a specific target plus enough structural context to judge layout quality:
- the screen or component name
- purpose of the page
- current UI code or screenshot
- device context
- content density constraints
- what feels wrong right now
- any immutable constraints such as existing design system tokens
Good targets:
- “pricing page hero and plan cards”
- “dashboard sidebar + content header”
- “mobile onboarding step layout”
- “settings form with dense controls”
How to turn a rough request into a usable prompt
Weak prompt:
- “Use arrange on this UI”
Better prompt:
- “Use arrange for the dashboard overview screen. The layout feels crowded and every card has the same visual weight. Keep the current components, but improve spacing rhythm, grouping, and hierarchy. Prioritize scanability on desktop, preserve the 12-column grid, and do not redesign colors.”
Why this is better:
- defines scope
- names the actual problem
- states constraints
- tells the model what can and cannot change
A practical arrange guide workflow
- Gather design context with
/frontend-design. - If the project lacks enough context, run
/teach-impeccable. - Point
arrangeat a single screen, section, or component set. - Ask for diagnosis before changes.
- Ask for revised layout decisions with reasons.
- Apply changes and review again with the same skill.
This sequence matters because arrange is strongest at structured critique, not blind guessing.
What arrange evaluates well
Based on the source, arrange is especially tuned to assess:
- spacing consistency versus arbitrary values
- whether everything has the same padding and therefore no rhythm
- whether related items are grouped tightly
- whether larger spaces separate groups
- whether the main focal point survives the “squint test”
- whether whitespace guides the eye
- whether the layout has an underlying structure
- whether repeated card grids are making the interface generic
If those are your pain points, the arrange skill is a strong fit.
Suggested prompt structure for better arrange usage
Use this input pattern:
Target:what exact screen or componentGoal:what should feel betterCurrent issue:what looks wrong nowConstraints:what must stay fixedContext:device, content type, design systemOutput wanted:critique, revised layout plan, code changes, or all three
Example:
Target: analytics dashboard main viewGoal: improve hierarchy and reduce visual samenessCurrent issue: every widget competes equally, spacing is uniform, sections blur togetherConstraints: keep existing Tailwind tokens and component inventoryContext: desktop-first SaaS appOutput wanted: diagnosis first, then layout revision recommendations
Repository files to read first
This skill is lightweight. The only meaningful file surfaced here is:
SKILL.md
Read it first because it contains:
- the dependency on
/frontend-design - the mandatory preparation step
- the layout assessment checklist
- the intended problem triggers
There are no visible helper scripts, references, or metadata files in this skill directory, so your install decision should be based mainly on how much you value the critique framework in SKILL.md.
What to ask arrange to return
To make arrange for UI Design practical, ask for output in one of these forms:
- a ranked list of layout problems
- a spacing and grouping plan
- a revised section structure
- before/after rationale
- implementation-ready CSS or component adjustments
If you only ask for “improve the layout,” you will usually get weaker, less actionable output.
Constraints that may block adoption
The main adoption blockers are straightforward:
- it depends on other skills in the same ecosystem
- it does not appear to ship examples, assets, or helper references
- it is guidance-heavy rather than tool-heavy
- it is strongest for critique and direction, not automated layout generation
If you want a self-contained skill with examples and implementation recipes, arrange may feel thin on its own.
arrange skill FAQ
Is arrange beginner-friendly?
Yes, if you already have a screen to improve. No, if you are starting from zero and need to learn UI design fundamentals first. The dependency on /frontend-design is a clue that the author expects broader design guidance to be available.
Is arrange better than a normal prompt?
Usually yes for layout-specific critique. A generic prompt often produces vague feedback like “add more whitespace” or “improve hierarchy.” arrange gives a clearer review lens: spacing rhythm, grouping, squint-test hierarchy, and grid monotony.
Can I use arrange without the rest of impeccable?
Possibly, but with reduced value. The skill explicitly requires /frontend-design and may require /teach-impeccable if no context exists. In practice, arrange install makes the most sense if you can access the surrounding impeccable workflow.
When should I not use arrange?
Skip arrange when your main issue is visual styling rather than layout structure. It will not be the best first tool for brand exploration, illustration direction, color systems, or deep accessibility remediation.
Does arrange work for code and screenshots?
It should work best when you provide either UI code, a screenshot, or both. Without concrete artifacts, the model has to imagine the layout, which lowers output quality.
Is arrange useful for design systems?
Yes, but mostly at the screen-composition level. It helps you apply existing tokens and components with better rhythm and hierarchy; it is not primarily a design-system authoring skill.
How to Improve arrange skill
Give arrange one layout problem at a time
arrange performs better on a single page region or screen than on an entire product. If you give it a full app, the advice tends to flatten into generic observations. Narrow scope produces sharper structural changes.
Show what must remain unchanged
High-quality arrange usage depends on constraints. Tell it what cannot move:
- grid system
- breakpoints
- component inventory
- token scale
- brand rules
- content order
This prevents unrealistic redesigns and keeps the suggestions implementation-ready.
Ask for diagnosis before redesign
A strong pattern is:
- “Audit the current layout.”
- “Rank the top 3 structural problems.”
- “Propose fixes with minimal changes.”
- “Only then suggest deeper alternatives.”
This improves trust because you can evaluate the reasoning before accepting changes.
Provide evidence of the current weakness
The best inputs mention symptoms, not just preferences:
- “users miss the primary CTA”
- “the cards read like one undifferentiated wall”
- “the form feels longer than it is”
- “sections compete for attention”
That gives arrange a real decision target instead of a vague aesthetics brief.
Common failure modes with arrange
Watch for these weak outputs:
- adding whitespace everywhere equally
- over-relying on card grids
- suggesting hierarchy through size only
- ignoring content relationships
- proposing changes that fight your design system
If you see these, ask the skill to rebalance grouping, contrast between dense and open areas, and section-level rhythm.
Stronger inputs for arrange for UI Design
Better than:
- “make this cleaner”
Use:
- “Improve the layout rhythm of this settings page. Related controls should feel grouped, advanced options should recede, and the save action should remain easy to find. Keep all fields and labels.”
This gives the skill a clear hierarchy and grouping objective.
Iterate on the first output the right way
After the first pass, do not just ask for “another version.” Instead ask:
- which spacing relationships changed and why
- what became the primary focal point
- how grouping improved scanability
- what tradeoffs were introduced
- how the layout should adapt on mobile
That forces more useful second-round reasoning.
Pair arrange with implementation requests
Once the critique is good, ask for the next layer:
- Tailwind class changes
- CSS spacing scale updates
- component restructuring
- grid adjustments by breakpoint
This turns arrange guide output into something a developer can ship.
Improve trust by asking for minimal-change and maximal-change options
A practical way to use arrange skill in teams is to ask for:
- a conservative revision
- a stronger structural rewrite
That helps stakeholders compare risk versus payoff without treating layout work as subjective guesswork.
