SaaS Security End — Auditoria de Segurança Pré-Lançamento
leonardoborges13/saas-security-end
Auditoria completa de segurança SaaS antes da produção com relatório detalhado
Ver no GitHubSkill do Claude Code que executa uma auditoria estruturada de segurança em 7 módulos antes do lançamento de uma SaaS. Cobre RLS, autenticação, pagamentos, frontend, infraestrutura, conformidade LGPD e checklist de testes manuais — identificando vulnerabilidades críticas e gerando correções automáticas.
Instalação
npx convez-ai leonardoborges13/saas-security-end1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Validar segurança de banco de dados (RLS) antes do lançamento
- Auditar rotas protegidas, autenticação e controle de acesso
- Verificar conformidade LGPD e proteção de dados sensíveis
- Testar integração com gateways de pagamento e webhooks
- Gerar checklist personalizado de testes pré-produção
Veredito

Ferramenta pragmática e bem-estruturada para reduzir riscos de segurança em lançamentos de SaaS, especialmente em stack Supabase + React. Requer projeto já em estágio avançado e adapta-se bem a variações arquiteturais.
README
A tradução para o português deste README ainda está na fila.
🛡️ saas-security-end — SaaS Security Sprint (Pre-Launch)
A battle-tested Claude Code skill that audits the security of any SaaS before it ships to production. Covers database (RLS), authentication, payments, frontend, infrastructure, LGPD compliance, and manual testing.
What it does
/saas-security-end runs a structured 7-module security audit on your SaaS project, designed to catch the issues that actually kill product launches:
| Module | Focus | Priority |
|---|---|---|
| 1. Database | Row-Level Security (RLS) on every table | 🔴 Critical |
| 2. Authentication | Route protection, admin checks, session safety | 🔴 Critical |
| 3. Payments | API keys, tokenization, webhook validation, server-side amounts | 🔴 Critical |
| 4. Frontend | Env vars, XSS, dependencies, console leaks | 🟡 Important |
| 5. Infrastructure | Security headers, CORS, HTTPS, backups | 🟡 Important |
| 6. LGPD / Data protection | Consent, ToS/Privacy pages, right to deletion | 🟡 Important |
| 7. Manual tests | Personalized pre-launch checklist | 🔴 Final gate |
For each item, Claude classifies as ✅ OK · ⚠️ Attention · ❌ Vulnerable — and fixes what it can automatically (frontend code, RLS SQL, env var moves).
Stack assumed
This skill was built against the most common Brazilian SaaS stack, but the principles apply broadly:
- Frontend: React (Vite or Next.js)
- Backend: Supabase (Postgres + Auth + Edge Functions + Storage)
- Payments: Asaas, Stripe, or any tokenizing gateway
- Deploy: Vercel, Netlify, or similar
- Compliance: LGPD (Brazilian GDPR equivalent)
Works on any project that matches this shape. Easy to adapt to other stacks (Firebase, Pocketbase, etc.) — the prompts are mostly stack-agnostic.
Installation
Option 1 — Clone directly into your Claude skills folder (recommended)
# macOS / Linux
git clone https://github.com/leonardoborges13/saas-security-end.git ~/.claude/skills/saas-security-end
# Windows (PowerShell)
git clone https://github.com/leonardoborges13/saas-security-end.git $env:USERPROFILE\.claude\skills\saas-security-end
That's it. Restart Claude Code and type /saas-security-end to invoke it.
Option 2 — Manual install
- Download
SKILL.mdandplaybook.mdfrom this repo - Create folder
~/.claude/skills/saas-security-end/ - Drop both files inside
- Restart Claude Code
Option 3 — Per-project install
If you want the skill only in one project:
mkdir -p .claude/skills/saas-security-end
curl -O https://raw.githubusercontent.com/leonardoborges13/saas-security-end/main/SKILL.md
curl -O https://raw.githubusercontent.com/leonardoborges13/saas-security-end/main/playbook.md
mv SKILL.md playbook.md .claude/skills/saas-security-end/
Usage
Inside any project directory in Claude Code:
/saas-security-end
Claude will:
- Read the playbook (
playbook.md) - Map your project — discovers tables, routes, Edge Functions, env vars, deploy platform, domains
- Run all 7 modules in order, adapting each prompt to your real code
- Fix what it can automatically and generate SQL/manual steps for the rest
- Generate a personalized manual test checklist with your real routes and domain names
- Output a final security report with health score and pending items
Typical run takes 15–40 minutes depending on project size.
When to run it
- ✅ Before launching a new SaaS to production
- ✅ When auditing an existing project's security posture
- ✅ When adding payments or sensitive data
- ✅ When switching from sandbox to production gateway
- ✅ Quarterly, as a regression check
What it does NOT cover
This skill targets low-to-medium-risk SaaS (under ~10k MAUs, no PHI, no PII at scale).
For advanced compliance, consult a specialist:
- PCI-DSS (handling raw card data — tokenization sidesteps this)
- SOC 2 (organizational + process controls beyond code)
- HIPAA (healthcare data)
- Penetration testing (red team / formal pentest)
The skill explicitly flags this limit in its final report.
Sample output
SECURITY REPORT — MyApp.com
Date: 2026-06-03
MODULE 1 — Database (RLS): 12/14 items OK
⚠️ Table `user_documents` has RLS enabled but no SELECT policy → blocks everything
❌ Table `payment_methods` has RLS DISABLED — CRITICAL, fix before launch
→ SQL fix generated in `migrations/sec-001-payment-methods-rls.sql`
MODULE 2 — Auth: 8/8 items OK ✅
MODULE 3 — Payments: 6/7 items OK
❌ Webhook endpoint not validating Asaas token → CRITICAL
→ Fix applied in `supabase/functions/payment-webhook/index.ts:42`
MODULE 4 — Frontend: 11/12 items OK
MODULE 5 — Infrastructure: 4/6 items OK
MODULE 6 — LGPD: 3/5 items OK
⚠️ No "delete my account" flow exists
⚠️ Privacy policy missing data retention section
VULNERABILITIES FOUND: 6
AUTO-FIXED: 3
PENDING (manual action): 3
SHIP READINESS: 🟡 BLOCKED on 2 critical items
Philosophy
Most SaaS security failures aren't sophisticated attacks — they're:
- A forgotten table without RLS
- An API key accidentally committed
- A webhook that accepts any payload
- A
plan_statusfield that the frontend can write to
This skill is a systematic sweep of the boring, high-impact failure modes. It won't find a zero-day. It will find the things that get exploited by anyone with browser DevTools and 5 minutes.
Contributing
PRs welcome. The most useful contributions:
- Stack adapters — versions for Firebase, Convex, Pocketbase, Appwrite
- Compliance modules — additions for GDPR, CCPA, HIPAA basics
- Gateway modules — specific checks for Stripe, MercadoPago, PagSeguro, dLocal
- Example reports — anonymized reports from real audits showing what the skill catches
Open an issue first for big changes so we can align on scope.
License
MIT — use it, fork it, sell consulting with it. Attribution appreciated but not required.
Credits
Built by Leonardo Borges.
Forged from real pre-launch audits. Each module exists because something went wrong on a real project.
"Para que possais executar a vossa salvação com temor e tremor." — Filipenses 2:12