Install
$ agentstack add skill-krishna-modi12-frontend-design-pro-web-interface ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
Web Interface Review
When to Use
Auditing or polishing existing UI rather than building new: design review, accessibility audit, copy review, typography and contrast passes, "make this feel more finished", "what's wrong with this component". This skill is mostly rules to check against rather than code to generate — but the craft rules have a worked example, cited below.
Stack
Framework-agnostic review · applies to React 19 · Tailwind v4 output
Core Rules
- Surface craft. Layered shadows (ambient + direct, ≥2 layers), semi-transparent borders, concentric nested radii, hue-consistent borders/shadows on tinted backgrounds.
- Interaction increases contrast. Hover, active and focus states must be more contrasted than rest — never less.
- Copy is UI. Active voice ("Install the CLI"), second person, sentence case outside marketing, specific button labels ("Save API Key", never "Continue"), errors that state the fix rather than the failure.
- Numbers and units. Numerals for counts ("8 deployments"), non-breaking space before units (
10 MB), consistent decimal places within a context, obviously-fake placeholders (YOUR_API_TOKEN_HERE). - Typography detail.
…not...· curly quotes · loading copy ends with an ellipsis ·tabular-numsin number columns ·text-wrap: balanceon headings. - Overflow is handled.
truncate/line-clamp/break-wordson text containers; flex children needmin-w-0or truncation silently fails. - Touch and safe areas.
touch-action: manipulation,overscroll-behavior: containin overlays,env(safe-area-inset-*)on full-bleed layouts, intentional-webkit-tap-highlight-color. - Images. Explicit
width/height(CLS),loading="lazy"below the fold,priorityabove it. - Rendering artifacts. Animate a wrapper rather than the text node (anti-aliasing shifts); use images not CSS gradients for long dark fades (banding).
- Locale.
Intl.DateTimeFormat/NumberFormat, never hardcoded formats;translate="no"on brand names and code tokens.
Patterns
- Audit output — group by file,
file:line — rule — one-line fix, terse, no preamble. - Contrast check — WCAG 2.2 AA is the gate; APCA (Lc ≥75 body) is the tiebreaker when a colour passes 4.5:1 but still reads poorly.
- Anti-pattern sweep —
user-scalable=no, blocked paste,transition: all,outline:nonewithout replacement, `, images without dimensions,autoFocus` without justification.
Live verification (rendered-DOM audit)
Every rule above checks source. No gate in this pack renders anything — a stylesheet that 404s, a 73px sideways scroll at 390px, a focus ring computing to 4.36:1, a reveal-on-scroll that died and left its text at opacity:0 all pass a green chain. When the ask is "audit the live site", "check how this actually looks", or "does this look right in the browser", verify the rendered page, not just the source.
- Hard dependency: Playwright MCP. Rendered-DOM inspection has no source-only fallback. If the
browser_*tools are not connected, say so and stop — never emit a source-only pass and call it a live audit. - Two layers, neither sufficient alone. The 61 constraints (
scripts/test_constraints.py) are Layer A; this rendered pass is Layer B. A page is verified only when both pass. Layer B never re-litigates what Layer A checks cleanly — it covers the render-only residue: computed contrast on arbitrary text, real overflow and overlap, dead reveals, console and network failures, focus order. - Engineering vs critique — never conflate them. Objective findings (broken layout, a11y failure, console/network error, measured contrast below AA) are reported as failures. Design critique (weak hierarchy, clutter, timid spacing) is a separate, labelled section that never fails the audit.
Workflow, the confirmed tool surface, the findings schema, source-mapping, and what still cannot be measured: references/live-verification.md.
Examples
examples/good-audited-panel.tsx is the worked craft pass: two shadow layers, hover states that gain contrast, tabular-nums on every numeric column, the min-w-0 + truncate pair, Intl formatting and translate="no" on identifiers — all four states, no faked delay. Read it when the ask is "make this feel more finished".
The anti-examples are the other half — what a review rejects: examples/bad-generic.tsx (AI-slop layout and copy) · examples/bad-inaccessible.tsx (a11y failures) · examples/bad-drive-by-refactoring.tsx (scope violations).
Reference Index
Load only for the specific task:
| Task | Load | |---|---| | Surface craft, copywriting, typography detail, safe areas, APCA | references/web-interface-guidelines.md | | 200+ granular UX rules with Apple HIG / Material citations | references/ux-deep-rules.md | | Core UX principles, animation rules, empty/error handling | references/ux-guidelines.md | | Live-browser rendered audit — workflow, tool surface, findings schema, source-mapping | references/live-verification.md |
Constraints
Report findings, don't silently rewrite — respect surgical scope (BEHAV-01). Any code you do produce still meets the full baseline: OKLCH tokens, WCAG 2.2 AA, four states, prefers-reduced-motion, TypeScript strict.
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: Krishna-Modi12
- Source: Krishna-Modi12/frontend-design-pro
- License: MIT
- Homepage: https://krishna-modi12.github.io/frontend-design-pro/
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.