improve-ui
/improve-ui
Krytykuje istniejący interfejs i audytuje powierzchnię produktu względem własnych dowodów projektowych, bez zmieniania kodu źródłowego.
- Autor
- aadarsh-jayswal ↗
- Repozytorium
- UI-UX-Master ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 28.10.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 1
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 28.10.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Krytykuje istniejący interfejs i audytuje powierzchnię produktu względem własnych dowodów projektowych, bez zmieniania kodu źródłowego.
Oryginalny opis (EN)
Critique an existing interface or audit a product surface against its own design evidence, then prioritize findings or prepare implementation plans. Read-only on product source; use when evaluation is requested rather than a clear direct correction.
Jak działa
- Modes
- Evidence for critique and audit
Podgląd SKILL.md
--- name: improve-ui description: Critique an existing interface or audit a product surface against its own design evidence, then prioritize findings or prepare implementation plans. Read-only on product source; use when evaluation is requested rather than a clear direct correction. --- # Improve UI Own independent review and principle explanation. Read [coordination](../../shared/references/coordination.md), [policy](../../shared/references/policy.md), the selected mode in [review procedure](../../shared/references/review.md), and the review or explanation section of the [quality bar](../../shared/references/quality-bar.md). ## Modes - **Critique:** Evaluate supplied rendered UI, screenshots, or a description with task/audience context. Identify heuristic recommendations as recommendations, not proven user behavior or measured defects. - **Audit:** Select one coherent surface, trace its runtime owners, and test candidates against its binding design evidence. Prefer no finding to an unsupported one. - **Plan:** Produce one self-contained [implementation plan](../../shared/templates/implementation-plan.md) per selected root correction, including exact scope, reuse, current evidence, expected outcome, and checks. - **Reconcile:** Re-open plan evidence and update status, stale locations, assumptions, and affected consumers. Do not implement the plan. - **Explain:** Answer a question about a UI principle or value directly: the correct ranges, the limits that make them actionable, a code example consistent with the advice, every applicable WCAG criterion with its number, level and threshold, and a quick test the reader can run. Use the [playbook](../../shared/references/playbook/index.md) entries to find the guidance, but cite public sources the reader can open, not playbook IDs, and say which statements are your own arithmetic. ## Evidence for critique and audit 1. Pick principles with the [task router](../../shared/references/playbook/task-router.md). When the input is code, check every declared value in its "CSS property → principles" table, not only values that look wrong. 2. Measure colour pairs with [contrast.mjs](../../shared/scripts/contrast.mjs) (`--non-text` for borders, rings and indicators at 3:1). 3. When the page renders, run [audit-rendered-ui.js](../../shared/scripts/audit-rendered-ui.js) through [page-eval.mjs](../../shared/scripts/page-eval.mjs) `--audit` at the stated desktop width and at `--width 390`; treat each result as a lead to confirm. Polish details (radius, shadows, icons, motion values) use [polish](../../shared/references/polish.md). 4. Report each finding with Before and After values and, where one applies, its public standard (WCAG criterion number and level) in the [findings format](../../shared/templates/findings.md), ordered by user impact, ending with the verdict and the not-verified list. Playbook IDs guide your selection; they do not appear in the report. Input is a bounded product surface and evidence. Source can prove token/variant/copy-contract contradictions; visual hierarchy or perceived clarity needs rendered/user evidence. Deduplicate root causes without merging distinct problems, prioritize by confidence and user impact, and default to at most three high-confidence findings for a design-drift audit unless broader coverage is requested; a critique or review of a whole screen lists every supported issue. Assess visual impact, brand identity, narrative, and interaction richness alongside clarity. Weight them by the brief rather than assuming restraint is superior. For redesigns, compare original imagery and actual hover/focus, scroll, and transition behavior; lost storytelling or animation can be a major regression. Use the animate review contract when motion is central. Static screenshots do not verify interaction quality, and visual richness alone does not prove a performance problem. Never change product source, install dependencies, run mutating formatters, or publish findings. Write only requested review/plan artifacts. Do not expand into functional debugging, architecture, performance, metadata, or a full accessibility audit merely because related code exists. A clearly requested small fix should go directly to its implementation owner instead. If implementation was already authorized, return the accepted scope to that owner without asking for the same authorization again. Otherwise finish with findings/plans and the choices that need product judgment. Done when every finding states its evidence type (measured, rendered, source or judgement), affected surface, supported correction, and verification. This skill remains separate to preserve the read-only assessment contract.
Wersje
| Repozytorium | Wynik | Licencja |
|---|---|---|
| Ta strona · aadarsh-jayswal/UI-UX-Master | 100 | MIT |
| ibelick/ui-skills | 100 | MIT |
| DanielD2G/dgstack | 100 | MIT |
| aadarsh-jayswal/UI-UX-Master | 100 | MIT |
Migawka bazy z 13.11.2026 · wersja paczki 95885dac98d5 (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 90audit-design-surface
Ocenia istniejący interfejs i ścieżkę użytkownika pod kątem użyteczności, stanów, dostępności, responsywności i spójności. - 90dx-design-critique
Ocenia istniejącą stronę produktu według katalogu standardów i wzorców, a potem zwraca uszeregowane sugestie poprawek bez zmieniania kodu. - 100design
Projektuje i krytykuje widoczne UI przed budową i po niej: wyznacza kierunek, ocenia powierzchnię, buduje i weryfikuje, czytając DESIGN.md i PRODUCT.md projektu. - 100superior-ui-audit
Przeprowadza audyt istniejącego interfejsu bez zmian w kodzie i przygotowuje rekomendacje oraz plany wdrożeniowe. - 100design-audit
Audytuje istniejący interfejs pod kątem użyteczności, hierarchii, dostępności i konwersji na podstawie adresu URL, zrzutu lub kodu, z priorytetyzowaną listą poprawek. - 90integra-design-qa
Przegląda istniejący interfejs lub przepływ pod kątem UX, jakości wizualnej, responsywności i ryzyk wdrożeniowych.