agent-diversity-review
CommandRun the Agent Diversity Review gate and emit the result table
One-line SaaS — from a description, scaffold a complete CF-native multi-tenant SaaS (Hono + D1 + Drizzle + Better Auth + Stripe + shadcn) deployed to a real URL
> /plugin marketplace add heymegabyte/agent-skills
description: One-line SaaS — from a description, scaffold a complete CF-native multi-tenant SaaS (Hono + D1 + Drizzle + Better Auth + Stripe + shadcn) deployed to a real URL argument-hint: <one-line-description-of-the-saas> allowed-tools: Read, Write, Edit, Bash, Glob, Grep, Agent, WebFetch, mcp__playwright__* model: "opus"
ULTRATHINK before responding. The user wants a complete SaaS: $ARGUMENTS
Per `[[02-goal-and-brief]]`:
Per `[[monitor-orchestration]]` + `[[source-site-enhancement]]` § Parallel-agent playbook. Spawn ALL in one multi-tool message:
| Agent | Role | |---|---| | `architect` | Project structure, file inventory, DO/D1/KV/R2/Vectorize decisions per `[[05-architecture-and-stack]]` | | `auth-implementer` | Better Auth or Clerk wiring with D1 schema, social providers, magic links, session storage | | `payments-implementer` | Stripe + Link (DEFAULT, all money flows; Billing for SaaS) / Square only on prompt request, per `[[payments]]`, webhook handlers, idempotency | | `feature-builder` (×3) | Per-domain features — one agent per domain (auth, billing, core feature) | | `frontend-builder` | React 19 + Vite + TanStack Router + Tailwind v4 + shadcn/ui; SSR via vite-ssg or TanStack Start | | `content-writer` | Marketing copy (Flesch≥60, active voice, anti-slop) per `[[copy-writing]]` | | `seo-auditor` | Per-route metadata, JSON-LD, sitemap, robots.txt with explicit AI-crawler allow/disallow | | `test-writer` | Failing Playwright specs BEFORE impl per `[[verification-loop]]` |
Per `[[cloudflare-lock-in-is-leverage]]`:
NO: Vercel, AWS, Supabase, PlanetScale, Inngest (unless explicit), Prisma. Per `[[cloudflare-lock-in-is-leverage]]`.
SaaS standard set: `/`, `/pricing`, `/features`, `/docs`, `/blog` + `/blog/{slug}`, `/about`, `/contact`, `/privacy` + `/terms` + `/dpa` + `/cookies`, `/login`, `/signup`, `/dashboard`, `/settings`, `/billing`.
SaaS jewels: `/changelog`, `/security`, `/status`, `/customers`, `/roadmap`, `/api`, `/sdk/{lang}`, `/compare/{competitor}`.
100 build-breaking rules:
1. Cinematic Visual Doctrine — dark-first, OKLCH, clamp() type 2. SPA + Routing — React 19, TanStack Router, View Transitions 3. PWA Kit — manifest with screenshots/shortcuts/share_target 4. Structured Data — 5+ JSON-LD per route (accurate, never padded) 5. SEO + GEO — title 50-60ch, desc 120-156ch, robots AI-crawler explicit 6. Integrations — Stripe/Square/Resend/Clerk/Turnstile/Workers AI all configured 7. Performance — LCP ≤2.0s, CLS ≤0.05, INP ≤100ms 8. A11y — WCAG 2.2 AA, axe 0 violations × 6bp 9. Testing — Playwright 6bp + Vitest + Lighthouse + AI vision QA ≥9/10 10. Deploy — atomic, CDN purge, /health, D1 backups
1. Provision via wrangler: `wrangler d1 create`, `wrangler kv:namespace create`, `wrangler r2 bucket create`, secrets from `[[secret-provisioning]]` 2. `wrangler deploy` → real `.workers.dev` URL or custom domain 3. Playwright on PROD URL — full donor/signup/checkout flow 4. AI vision QA on every new route 5. Lighthouse CI all green
Per `[[always]]` § Hard Gates and `[[16-cinematic-website-prime-directive]]` Done = deployed at a real URL + every Hard Gate green + Self-Verify Statement per route + announced.
Include: deployed URL, route inventory (count), validation gate matrix (pass/fail per gate), token spend, model trace per `[[model-routing]]`.
Agent Skills — agent-neutral autonomous product-building OS for 32+ AI coding tools. 23 skill categories · 28 agents · 149 reference docs · 37 platform variants. One-line prompts → deployed products on Cloudflare Workers.
Run the Agent Diversity Review gate and emit the result table
Report the Claude control-plane environment (auth mode, tooling, connectors, resolved target). Fires when you need to confirm the…
Bring ANY repo into compliance with the shared Claude control plane so Claude Code on the web can run /ship autonomously.
Reconcile the Cloudflare agent fabric: official skills, API MCP, MCP Portal candidates, AI Search, Secrets Store, Dynamic Routes,…
Scaffold a new spawnable specialist agent def and register it in the agent taxonomy
Audit whether user-visible changes in the current session have matching CHANGELOG.md entries; report MISSING with suggested…
Implement tasks incrementally — build, test, verify, commit. Add "auto" to run the whole plan in one approved pass.