Open-source verification stack for AI-assisted TypeScript apps (92-rule security scanner, headless Playwright QA & 72-rule UI linter)

[avivgerz-hue]

Hey everyone! I built three free, open-source TypeScript CLI & MCP tools to catch the security, runtime, and UI issues AI coding agents leave behind:

1. vibe-shield — 92-rule static security scanner for TypeScript, Hono/Next.js/Node API routes, Supabase, Drizzle/Prisma, and MCP/Skill files (SKILL.md, .cursorrules, mcp.json):
   npx -y @vibefoundry/vibe-shield@latest scan .
   Repo: https://github.com/avivgerz-hue/vibe-shield

2. ghost-qa — Zero-config headless Playwright QA engine that tests your live local dev server (http://localhost:3000 / http://localhost:5173) across 8 runtime traps (console/page crashes, failed 4xx/5xx network requests, DOM collisions, accessibility, Core Web Vitals, broken links, SEO, and dark/light parity) on Desktop, Tablet, and Mobile:
   npx -y ghost-qa test http://localhost:3000
   Repo: https://github.com/avivgerz-hue/ghost-qa

3. unslop — 72-rule static UI & CSS design craft linter (WCAG 2.1 AA/AAA + APCA contrast math, typography hierarchy, spatial grid rhythm, and 10 studio design archetypes):
   npx -y unslop-mcp audit .
   Repo: https://github.com/avivgerz-hue/unslop

All three run locally in seconds with npx (or bunx) and also work as MCP servers and Agent Skills (SKILL.md). Feedback and edge-case reports are very welcome!