ZipLoom
Home
PricingSign InGet Your Live URL
10Checks On Every Deploy

security

Security Scan On Every Deploy

Catch RLS Gaps, Secret Leaks, And CVEs Before Your App Is Reachable.

Updated

At a glance

ZipLoom scans every deploy for missing RLS, leaked secrets, known CVEs and license risk — the kind of gap that leaked 1.5M tokens in production.

  • Row-Level Security Enforcement
  • Secrets Detection
  • CVE And License Audit
Sample. ZipLoom scans the exact commit you deploy and stops the security mistakes AI coding tools make before they ship. A critical finding, like a live key pasted into source, holds the deploy until a fixed commit passes the scan.

Why it matters

AI coding tools write working code quickly, and they repeat the same handful of mistakes: a database table left readable by anyone, a key pasted into a source file, a dependency with a published vulnerability, a route that never asks who is calling. None of these break the app. They pass every manual click-through, and that is why they reach production.

ZipLoom puts the check in the one place nobody can skip: the deploy. The exact commit being shipped is read before the build starts, a critical finding stops the release, and every finding comes back with what is wrong, where it is, and how to fix it. Findings your team has already triaged stay triaged, so the gate keeps meaning something instead of becoming a step people learn to click past.

What Happens Before Your Code ShipsIllustration · sample data
Commita1b2c3dCodeDependenciesDatabase rulesConfigurationgatecritical: stoppedBuild + deployWhat · where · fixper finding

Real incident

The Incident This Prevents

CVE-2025-48757: a researcher scanned 1,645 apps built with one AI tool (Lovable) and found 170 with publicly readable databases. The root cause was a single unchecked Postgres setting — row-level security was never switched on. ZipLoom checks for exactly that setting before your first deploy. The tool's vendor disputes that the CVE describes a vulnerability in its product.

Source: Matt Palmer's published scan · CVE-2025-48757

What you get

Row-Level Security Enforcement

ZipLoom checks that RLS is enabled on every Postgres table before your app goes live. CVE-2025-48757 (a researcher found 170 of 1,645 apps built with one AI tool with publicly readable databases) and the Moltbook breach (1.5M authentication tokens) both traced to this single unchecked setting.

Secrets Detection

Scans for API keys, database URLs, and auth tokens in your codebase and environment config before deploy. Credentials in source code is the most common initial access vector in cloud breaches.

CVE And License Audit

Checks your dependency graph against known CVEs. Also flags dependencies with GPL or AGPL licenses that could affect your commercial distribution rights — a real issue for SaaS products built on open-source stacks.

Access And Input Checks That Flag

The code scan also flags webhook endpoints that never verify the sender, data routes that never check whose record is being requested, forms that let a caller set fields they should not, and code that rebuilds objects from untrusted input. These are heuristic: they flag what deserves a human look and will not catch everything.

Zero Configuration

No YAML files, no scan configuration, no custom rules to maintain. The scanner runs automatically on every deploy based on your detected framework and package manager.

How It Works

  1. 1

    You push your code or trigger a deploy from your dashboard.

  2. 2

    ZipLoom reads the exact commit being deployed — source files, CI workflows and lockfiles.

  3. 3

    The checks run before the build: a code scan (injection, XSS, missing auth, unsafe CORS, secrets in 10+ provider formats), RLS verification on your Supabase migrations, CVE and malware lookup against the OSV database, license classification of your dependency tree, and CI workflow checks. A critical finding stops the deploy — unless your team has already triaged it.

  4. 4

    Results appear in your deploy dashboard before traffic is routed to the new version.

  5. 5

    If you've set up the GitHub Actions integration, a failed scan blocks the merge — not just the deploy.

What this check inspects

  • Server-side code paths where user input reaches a database query, a file path, a shell command, or a template — the injection classes AI tools reproduce most often.

  • Whether routes that read or write user-owned data appear to check who is asking and whose record it is — including routes an AI added late and never wired to a guard. A heuristic check that flags what looks unguarded, not a proof that every route is safe.

  • Webhook handlers that act on a call without verifying who sent it, request bodies passed straight into a record update, and objects rebuilt from untrusted input.

  • Client bundles for credentials, internal URLs, and debug endpoints that were meant to stay server-side.

  • Framework configuration — the defaults that are safe in development and dangerous in production.

What it means when this fails

A critical finding means a stranger can reach something you intended to be private: a query they can alter, a record belonging to another account, or a key that unlocks a paid API. ZipLoom blocks the deploy rather than shipping it and telling you afterwards.

What it does not do

  • It does not replace a penetration test. A person probing your business logic finds things a mechanical scan cannot.

  • The access and input checks are heuristic. They flag what deserves a human look; they do not prove every route is safe.

  • It does not change your code. It reports the fix; you or your AI tool apply it, and Fix It For Me can open the change as a pull request for you to review.

Questions

Does this replace a penetration test?
No. A pentest is a human expert probing your specific business logic. This catches the mechanical, repeatable mistakes AI tools make — which is most of what goes wrong, but not all of it. Treat it as the floor, not the ceiling.
Will it block my deploy over something trivial?
Only critical and high findings gate a deploy. Lower-severity findings are reported with the fix and let the release through, so the gate stays meaningful rather than something you learn to bypass.
Does my code leave my repo?
The scan runs against your repository and your deployed app inside ZipLoom's pipeline. Findings are stored with your organization; your source is not used to train anything.

What it replaces

Snyk Team

Per contributing developer, 5-seat minimum (~$1,500/yr entry). Code + dependency scanning; does not deploy your app.

$25/dev/mo

GitHub Code Security

CodeQL scanning, Dependabot, dependency review. Billed per active committer; GitHub repos only.

$30/committer/mo

Semgrep Team

SAST with a strong free OSS engine; rules and triage are yours to maintain.

$35/contributor/mo

ZipLoom Solo — every deploy check included

Deploy, security scan, and monitoring in one flat price. Guild $40/mo · Studio $70/mo · LoomWorks $149/mo · Business $299/mo.

$216/yr($20/mo)

Related features