# Produced Skill > Test fixture — produced SKILL.md ships a complete Setup (root-entry-file + Companion CSS) but Hard rules has zero lines matching shell vocabulary AND a token shape. SHELL_INVARIANTS must FAIL with the count assertion message naming the three pre-seeded slugs. - Skill: `vercel-labs/produced-skill-14` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add vercel-labs/produced-skill-14` - Raw SKILL.md: https://api.skillmd.com/api/skills/vercel-labs/produced-skill-14/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Web & Frontend - Author: Vercel Labs (https://skillmd.com/u/vercel-labs) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/vercel-labs/produced-skill-14 --- ## Setup Source: `` @ `src/app/layout.tsx` ```tsx import { ThemeProvider, BaseSurface } from 'ds-react' import './globals.css' export default function RootLayout({ children }) { return ( {children} ) } ``` ### Companion CSS — src/app/globals.css ```css @import "ds-themes/light.css"; body { background-color: var(--ds-surface-default); color: var(--ds-text-default); } ``` ## When to Load References | Trigger | Files to load | Notes | |---|---|---| | button asks | references/components/button.md | per-component file | ## Hard rules - Any prop, variant, token, or asset the agent cannot ground in source gets a literal `[VERIFY]` marker inline. - Mark unverifiable facts `[VERIFY]`. Report blockers instead of guessing. ## Final checks After generating UI: cite each component used to its source file; list any `[VERIFY]` markers; name the screen-level prompt built. In scope: tokens, assets, component descriptions, component APIs.