design-ui
Skill0 starsChanged 3 days agoMITView on GitHub
Design and iterate a reviewable UI from the user's jobs, using the project's design authority and existing screens, to the shared UI bar (jobs that finish, one owner per feature, simple, scannable, one component system, fast). Use for UI design requests or a workflow's design/prototype handoff; a direct build request can use the working UI as its prototype.
12 other projects use a version of this.
Adopted by
Most recently taken up by
- adelrodriguez/skillstheir own version, first seen Nov 15, 2026
- cachewraith/cachewraith-skillstheir own version, first seen Oct 25, 2026
- melconcoast/code-ideatheir own version, first seen Oct 13, 2026
- kdy1/kdy1-scriptstheir own version, first seen Oct 13, 2026
- NachikethReddyY/skillstheir own version, first seen Oct 13, 2026
- mae616/ai-templatetheir own version, first seen Aug 22, 2026
Used with
Projects with it also use testing.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
NVIDIASkilltesting
Run Python and Rust tests for physicsnemo-curator using uv + pytest with coverage reporting via pytest-cov, and cargo-nextest for Rust tests. Includes benchmark workflows with pytest-benchmark and criterion.
Also in 3 projects that use this
Other versions12 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 12
- ThisIsSadeghi/KMPilot★ 56same name
- kdy1/kdy1-scripts★ 34same name
- mae616/ai-template★ 9same name
- melconcoast/code-idea★ 4same name
- shreyas-makes/agent-skills★ 2same name
- adelrodriguez/skills★ 1same name
- evgenii-studitskikh/Claude-Code-SaaS-Studio★ 1same name
- leolionart/claude-skills★ 1same name
- NachikethReddyY/skills★ 1same name
- ThisIsSadeghi/Loupe★ 1same name
2 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository30 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- llms.txtllms.txt
- agent-first-docsSkill
- analytics-reviewSkill
- architectSkill
- autopilotSkill
- browseSkill
- create-prSkill
- foremanSkill
- harness-auditSkill
- humanizer-viSkill
- image-genSkill
- implementSkill
- investigateSkill
- marketing-videoSkill
- perf-auditSkill
- plan-draftSkill
- qaSkill
- reconSkill
- release-prdSkill
- releaseSkill
- review-prSkill
- semantic-decisionSkill
- setup-projectSkill
- teamSkill
- telemetry-reviewSkill
- testSkill
- token-auditSkill
- triageSkill
- ui-auditSkill
The file
We split this skill into the rules it gives an agent
8 rules. 0 are common, 5 are distinctive, and 3 are very specific to this project.
0
Common
Most of the 10 edited versions we read keep these.
5
Distinctive
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 13
Follow preamble and the Auto-Decision Framework.
- Distinctive: line 19
Read the brief (conversation or ticket), AGENTS.md and its linked architecture and design docs.
- Very specific: line 31
Document reusable project decisions only when needed, in the existing design authority or a concise DESIGN.md linked from AGENTS.md.
- Distinctive: line 39
Jobs: Write each job the change serves as trigger → done with its ideal steps, as rows of the flow inventory (create the file with these …
- Distinctive: line 49
Lifecycle: Every state the backend can put the object in has a place and an exit (J3).
- Distinctive: line 56
When the design authority is silent on something this depends on (who owns a fact, which component is the loading state, the page gutter), …
- Very specific: line 71
Use browse or the caller's browser tooling: render, inspect screenshots at the common and the narrowest supported width, fix what falls …
- Distinctive: line 110
Do not commit, push or deploy as part of this skill.
“Common” is measured: kept by at least 6 in 10 of the 10 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
design-ui
Make it really good, use lots of tokens and iterate till you're proud of it.
Own project continuity and a reviewable design handoff. Visual direction comes from the brief and the design authority, not a preset style. Usability is not taste: the design meets the UI bar, whose row ids (J1, O1, C2 …) are cited below.
Follow preamble and the
Auto-Decision Framework.
Shared refs are filesystem paths beside this skill's directory, so read them
by path, not as skill://.
Project continuity
Read the brief (conversation or ticket), AGENTS.md and its linked architecture
and design docs. The declared design doc is authoritative wherever it lives.
Inspect actual components, tokens and the nearest screen; extending a page
means reviewing it in context and inheriting its local patterns. Read the
flow inventory beside the design authority (e.g. docs/ux/flows.md) and the
latest ui-audit report when one exists (ticket dir or .babysit/ui-audit/):
its open findings on the screens you touch are part of the brief.
Preserve the project's stack, library, content and asset constraints. For a new product, derive a direction from the audience and purpose; missing brand rules allow design judgment. Do not invent business facts or rebrand an existing UI.
Document reusable project decisions only when needed, in the existing design
authority or a concise DESIGN.md linked from AGENTS.md. No mandatory schema
or exhaustive inventory.
Fit the product before drawing
A brief asks for an addition, and a product designed one addition at a time ends with the same number on three screens and one job spread over four. Settle these before any layout:
- Jobs. Write each job the change serves as trigger → done with its ideal steps, as rows of the flow inventory (create the file with these rows when it is missing). Design the path first, then the screens it needs (J1, J2).
- Owner. For each fact and action in the brief, find the screen that already owns it or its nearest relative, and extend that. A new screen, tab or navigation entry needs a job no existing screen can own, and names what it replaces (O1, O2).
- Subtract. List what the change makes redundant (a duplicate summary, a tab, a setting, a step) and remove it in the same design. "Nothing to remove" is stated with its reason.
- Lifecycle. Every state the backend can put the object in has a place and an exit (J3); read the states from the code, not the brief.
- System. Name the shared component for each job on the screen: page shell, list row, form footer, sheet, loading, empty, error (C1). No fit → propose a shared component or variant as a decision; never a local one-off (C4).
When the design authority is silent on something this depends on (who owns a fact, which component is the loading state, the page gutter), decide it and record it there in a line or two. That is what stops the next change drifting.
Make and iterate
Deliver a preview the user can open. A design-only phase keeps it isolated from production navigation/state; a direct build request uses the requested app, without a duplicate throwaway UI. Show existing-page changes in context.
Each changed screen is shown loaded, loading, empty, failed and partly failed, built from the named shared components (C2), with production-shaped data: the longest real list, a missing value, an 80-character name, a large number (D1–D3). Never lorem or three tidy rows.
Use browse or the caller's browser tooling: render, inspect screenshots at
the common and the narrowest supported width, fix what falls short, and
recheck. On the rendered preview:
- Walk each job, counting actions and screen switches against its ideal steps (J2, J4), plus one relevant failure.
- Cold read (S2): a fresh subagent sees only the screenshot of each new or reshaped screen and answers "what is this for?" and "what would you do first?". No subagent tool → report S2 as unchecked.
- Count per changed screen: primary actions, tabs, row actions, rows over 2 lines, helper paragraphs, repeated facts (S1, S3, S4, T1–T3). Prefer layout, position and icons to sentences.
- Lay it beside its two nearest sibling screens: header, gutter, rows and states match (C1–C3).
- Name what the screen fetches on open and what it shows before data lands (P1, P4).
- The other color scheme and
agent-browser a11yon each screen.
ui-audit's checklists say how each
row is measured on the built UI. Honor requested build/verification commands.
A build is not visual evidence; report blocked inspection. Keep previews local
unless publishing is requested.
Handoff
Record decisions the UI cannot convey, deliberate departures from project
rules, unresolved facts, links to the preview/verification evidence, and the
UX contract implement builds to and qa / ui-audit check the build
against:
Jobs: <flow id: trigger → done · ideal steps · steps in this design>
Owner: <fact or action → owner screen> · removed or replaced: <what, or the reason for nothing>
Components: <job → shared component> · new: <name, why none fits, where recorded>
States: <screen → loading / empty / error / partial components>
Bar: <rows checked on the preview with their count or shot · rows unmet and why>With a ticket, resolve bbs ticket path design --write, write the note there,
then bbs ticket set-pointer design <path>. Otherwise an inline note suffices.
The caller owns review checkpoints; neither add approval rounds nor skip held ones. Preserve the reviewed UI as the implementation baseline for comparison. Do not commit, push or deploy as part of this skill.
STATUS: DONE | DONE_WITH_CONCERNS | NEEDS_CONTEXT | BLOCKED
VERDICT: DESIGNED
PROTOTYPE: <path/URL and how to open; or why unavailable>
VERIFY: <rendered views, walked jobs with step counts, bar rows checked and gaps>
SUMMARY: <what was removed or merged, project decisions and design note location>
NEXT: <caller's next step or unresolved prerequisite>A text-only spec is not a completed visual design.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

