libra-test

Run static rules against your codebase based on your Cursor and Claude rules.

Usage

Run rules from your repo’s .libra/rules/ — runs every .rule.js, .rule.ts, and .rule.sh in that folder:

npx @libra-mcp/libra-test

Or use a built-in template (no setup):

npx @libra-mcp/libra-test --template baseline-ui

Multiple templates:

npx @libra-mcp/libra-test --template baseline-ui --template fixing-accessibility --template fixing-metadata

Options: --template <name> (repeatable), --strict (exit 1 on warnings too), --json, --list (show templates).

What it does

• Project mode (no --template): Looks for a .libra/rules/ folder in the current directory. If present, runs every .rule.js, .rule.ts, and .rule.sh in it (.rule.ts is run via tsx); .rule.md files are counted as LLM rules (skipped in CLI). If the folder is missing, you’ll get a short message and exit code.
• Template mode (--template <name>): Runs predefined static rules from a built-in template (e.g. baseline-ui) against the current working directory.
• Output: File-specific violations with paths and line numbers. Failures (✗) and warnings (⚠) are listed separately. LLM rules are counted and skipped with a pointer to libra-mcp.com.
• Severity: Rules that say “unless requested” (e.g. no gradient, no tracking) use warn severity: they are reported under “Warnings” and do not set a non-zero exit code. Use --strict to make warnings fail the run (e.g. in CI).
• Exit code: Non-zero if any static rule failed (errors). Warnings alone do not fail; add --strict to fail on warnings too.

Templates

• baseline-ui — Opinionated UI baseline (Tailwind, accessibility, layout). From ui-skills baseline-ui. Includes 6 static rules (h-dvh, icon button aria-label, no arbitrary z-index, 200ms interaction feedback, no gradient/tracking unless requested) and 8 LLM rules (skipped in CLI; run via GitHub App).
• 12-principles-of-animation — Audit animation quality using Disney's 12 principles adapted for web (timing, easing, physics, staging), with file:line findings format. Includes 3 static rules and 3 LLM rules.
• fixing-accessibility — Audit and fix HTML accessibility (ARIA labels, keyboard, focus, forms, contrast, motion). Includes 4 static rules (icon-button aria-label, no div/span as button, no tabindex > 0, empty link) and 9 LLM rules.
• interaction-design — Design and implement microinteractions, transitions, loading states, gesture interactions, and feedback patterns with accessibility and performance guidance. Includes 2 static rules and 3 LLM rules.
• fixing-metadata — Audit and fix HTML/SEO metadata (titles, descriptions, canonical, OG, Twitter cards, favicons, JSON-LD). Includes 4 static rules (no duplicate title, html has lang, viewport meta, no duplicate meta description) and 8 LLM rules.
• fixing-motion-performance — Audit and fix animation performance issues (layout thrashing, compositor properties, scroll-linked motion, blur/filter usage, and tool boundaries). Includes 4 static rules (no layout transition, no scroll-event animation, no will-change layout, animation requires reduced-motion) and 4 LLM rules.
• swiftui-ui-patterns — Best practices and examples for SwiftUI view composition, state ownership, environment injection, TabView/NavigationStack wiring, and sheet patterns. Includes 2 static rules and 3 LLM rules.
• tailwind-patterns — Production-ready Tailwind CSS component patterns for layouts, cards, navigation, forms, buttons, typography, responsive behavior, dark mode, and correction rules for common mistakes. Includes 5 static rules and 3 LLM rules.
• ui-ux-pro-max — Comprehensive UI/UX design intelligence for accessibility, interaction, performance, responsive layout, typography/color, animation, style selection, and chart guidance with stack/domain search workflows. Includes 2 static rules and 4 LLM rules.
• wcag-audit-patterns — WCAG 2.2 accessibility audit checklist and remediation guidance across POUR principles, automated testing, and common fix patterns. Includes 6 static rules and 4 LLM rules.