Code audit services for apps built fast on Supabase and Postgres
A code audit is a manual review of an application's source code and database for security holes and data leaks. Goldenroz audits web and mobile apps built quickly with Lovable, Bolt, Cursor or Claude Code on Supabase and PostgreSQL. You get a findings report ranked by severity, fixes delivered as pull requests, and an automated daily probe left running. The price is fixed and quoted after a 30-minute scoping call. It is not a certified penetration test or a compliance attestation.
Fixed price, quoted after a 30-minute scoping call
Updated 2026-11-13. By Shai Goldenberg

Who this is for
- Founders whose vibe coded app now has paying users and real personal data.
- CTOs who inherited a codebase written largely by AI tools and do not know what is exposed.
- Teams on Supabase that are unsure every table has correct Row Level Security.
- Companies preparing for a customer security questionnaire, a funding round or an acquisition review.
Who it is not for
- Companies that need a certified penetration test report for a contract. Hire an accredited testing firm.
- SOC 2, ISO 27001, HIPAA or PCI attestation. We are not an audit or certification body.
- Network, cloud infrastructure or endpoint security reviews.
- Red-team exercises or social engineering tests.
What is included
- Database access reviewEvery table, view and storage bucket checked for Row Level Security: enabled or not, and whether each policy grants what it should.
- Anonymous and cross-tenant testsCalls made as an anonymous user and as a logged-in user of another account, to see what data and functions answer.
- Function and API reviewDatabase functions, edge functions and API routes checked for missing authorization, privilege escalation and unsafe input handling.
- Secrets and keysA search of the repository, the client bundle and the git history for service keys and credentials that should never reach a browser.
- Authentication and payments pathsSign-up, login, password reset, role changes and payment webhooks reviewed for bypasses and replay.
- Findings reportEach finding with severity, evidence, how to reproduce it and the fix, written for both engineers and executives.
- Fixes as pull requestsFixes for the findings arrive as pull requests in your repository, for your team to review and merge.
- Automated daily probeA scheduled check left running that calls sensitive functions as an anonymous user every day and flags anything that opens up.
How the work runs
Scoping call
30 minutesWe cover the stack, the size of the codebase and the data at risk. You get a fixed price and a delivery date.
Access and setup
Step 1Read access to the repository and a staging or read-only database connection. Testing is non-destructive.
Audit
Step 2Manual review of schema, policies, functions and code, plus live tests as anonymous and cross-tenant users. Critical findings are reported the day they are found.
Report and fixes
Step 3The findings report is delivered and walked through on a call. Fixes follow as pull requests, most severe first.
Retest and probe
Step 4Merged fixes are retested, and the daily probe is installed in your CI.
Vibe coding security: what AI-built apps get wrong
Tools like Lovable, Bolt, Cursor and Claude Code produce a working app in days. They optimize for the feature the prompt asked for. Access control is rarely in the prompt, so it is often missing, partial or wrong.
The pattern repeats across vibe coded apps. A table is created without Row Level Security. A policy checks that a user is logged in but not that the row belongs to them. A database function runs with owner rights and trusts an ID sent by the client. A service key ends up in frontend code.
None of this shows in a demo, because the app works for the person testing it. It shows when someone opens the browser console, copies the public key and asks the API for every row. A security code review looks for exactly that path.
Supabase security: RLS, functions and keys
In Supabase, the browser talks to the database through an API using a public key. That is by design. It means Supabase RLS policies are the only thing between an anonymous visitor and your tables. A table with RLS off, or with a policy that returns true, is readable by anyone who has the key.
Functions are the second gap. PostgreSQL lets every role execute a new function by default unless that right is revoked, and functions marked as security definer run with their owner's rights. One such function without an authorization check can skip every policy you wrote.
We check each table, policy, function and bucket by hand and by live calls. In multi-tenant apps we also test that one tenant cannot read or write another tenant's rows, and we recommend enforcing isolation in database triggers where policies alone are easy to get wrong.
Code audit, software audit and penetration test: which one you need
A code audit reads the source and the database with full access. It finds flaws an outside tester might never reach, and it produces fixes. Web application penetration testing attacks the running system from outside, usually by an accredited firm, and produces a report that customers and regulators accept.
Our service is the first kind: an application security assessment of code and database, with fixes. It is not a certified penetration test and it does not result in a compliance attestation. Many teams do our audit first, fix what it finds, and then commission the formal test with fewer surprises.
A wider software audit of architecture and code quality can be added to the scope. Larger rework that comes out of an audit is scoped separately as a build phase. If AI features are planned, see AI integration for how model access to data is controlled.
What backs this up
- Row Level Security at scaleShai maintains a production platform with about 145 tables under Row Level Security and 155 policies, handling real payments.
- Tenant isolation in the databaseOn that platform, tenant isolation is enforced in database triggers, in addition to policies.
- The daily probe is his own toolHe runs a daily automated probe that calls sensitive functions as an anonymous user and closes open permissions. The same approach is installed for you.
- Guard hooks from real incidentsHis repositories use guard hooks that block two SQL patterns that caused real incidents.
- A named engineer auditsShai, our CTO, does the review and writes the fixes. The report, the pull requests and the probe are yours.
What it costs
The audit is a fixed price, set after a 30-minute scoping call. The price depends on the size of the codebase, the number of tables and functions, and how many apps share the database.
- Code and database security auditQuoted after a 30-minute scoping callFixed price
- Fixes as pull requestsFor the findings in the reportIncluded
- Daily probe setupRuns in your own CIIncluded
- Scoping call30 minutes$0
Who does the work
Shai leads the technology side: business software, SaaS and mobile apps, payments and escrow systems, Postgres security, AI agents built with Claude Code and MCP, and IT and infrastructure projects with senior infrastructure engineers.
Questions buyers ask
What do code audit services include?
At Goldenroz: a manual review of your code, database schema, access policies and functions, live tests as anonymous and cross-tenant users, a findings report ranked by severity, fixes as pull requests and an automated daily probe. The scope is the application and its database.
How much does a security audit cost?
It is a fixed price quoted after a 30-minute scoping call. The price depends on the size of the codebase and the database. We do not publish a single number because a small app and a multi-tenant platform differ widely in scope.
Is this a penetration test?
No. It is a code and database security audit done with full access to the source. It is not a certified penetration test and does not produce a SOC 2, ISO 27001, HIPAA or PCI attestation. If a contract requires one of those, you need an accredited firm.
Is my vibe coded app secure?
It depends on the access rules, and those are the part AI builders most often get wrong. Common problems are tables without Row Level Security, policies that do not check row ownership, and service keys in frontend code. An audit tests each of these directly.
Do you only audit Supabase apps?
Supabase and PostgreSQL are where our evidence is strongest. We also audit TypeScript and Node.js applications on other PostgreSQL setups. The scoping call confirms whether your stack is a fit.
Do you fix the problems or only report them?
Both. Fixes for the findings are delivered as pull requests that your team reviews and merges, most severe first. Larger redesigns are scoped separately.
Will the audit affect production?
Testing is non-destructive and runs against staging where one exists. Where production must be checked, we use read-only calls and agree the test window with you in advance.
Security audit: by industry and use case
Guides on this topic
- The five security failures we find in apps built fast with Lovable, Bolt, Cursor or Claude Code on Supabase, plus a self-check list. Run it before launch.
- MCP is an open standard that connects AI applications to your tools and data. Learn how hosts, clients and MCP servers work, and what to secure first.
- Seven Claude Code practices from a production codebase: CLAUDE.md rules, guard hooks, worktrees, permission envelopes and CI review. Apply them to your team.
Related services
- Multi-tenant SaaS products with auth, billing and database security in the first release.
- Connect ERP, CRM, SQL and Microsoft 365 to Claude or ChatGPT through MCP, read-only and audited.
- CI/CD, release engineering, staging and production separation, observability and Docker.
- A 4 to 8 week program that sets up Claude Code in your repo with rules, hooks and a playbook.
Talk to the person who will do the work
One hour, no sales team. You talk to Shai Goldenberg, the CTO, about your systems or your product, and you leave with a clear next step and a price range.