#009Learning3 min
Security risks in apps built with Bolt, Replit and Lovable
Review generated code, exposed secrets and database permissions without blaming the builder.
Hackaiz team · updated 21/11/2026

01Identify where each part of the application lives
Bolt, Replit and Lovable can be part of an application’s creation workflow, but the editor name is not a security finding. Start by distinguishing the tool used to generate code from the place where the website is published and the service holding its data. A project edited in one tool can use a custom domain, a GitHub repository and a separate database. Those components have different access rules and require different evidence.
Before requesting an analysis, list the published address, the repository that matches the deployed code and the database used by that deployment. Keep test and production addresses clearly separated in your own notes. Hackaiz provides dedicated information pages for these builders and a common Connect workflow. It does not require you to pretend that the builder itself hosts every part of your application or controls the DNS of a domain you purchased elsewhere.
02Examine secrets and authorization, not the origin of the code
A useful code review asks whether privileged credentials have ended up in files delivered to a browser and whether permissions match the intended user roles. For a Supabase-backed application, a service key in frontend code and a table without appropriate row-level protection are concrete issues to investigate. Row-level protection means deciding which records a user may access, not merely hiding a page or a button in the interface. These are examples of risks, not assertions about your project.
To inspect source code, export or synchronize it to GitHub if your builder supports that workflow, then connect through the Hackaiz GitHub App and choose an allowed repository. Do not assume a deployed URL reveals private source files. Use a separate Supabase connection when you want database evidence. For live PostgreSQL or MySQL, the database audit examines accounts, rights and server settings through a read-only connection rather than reading application table contents.
03Authorize the exact deployment you want tested
Every website and API analysis requires proof of domain control. On Connect, enter the published address and choose an available proof method. For builder-managed addresses, the interface recommends the verification file because the builder controls that address’s DNS. Use “Copy the instruction for my AI” to ask the project’s AI to create the file, publish the project, and check that /.well-known/hackaiz-verification.txt serves the expected value.
If you control a custom domain’s DNS, you can instead publish the displayed TXT record under _hackaiz for the relevant hostname. Click “Verify now”, then record the separate authorization to analyze that domain without destructive tests. A change of hostname is not a reason to reuse proof blindly. When selecting components and checks, remember that unselected checks do not run and are not billed. Requested actions use credits; the scope should reflect the deployment you actually intend to examine.
04Review changes instead of trusting a regenerated page
When reading results, use Danger, To fix and To watch together with the evidence and coverage limits. For a database issue, identify the affected object and role; for a code issue, examine the relevant source context. A page that now looks correct does not by itself show that its underlying access rule changed. Likewise, an analysis with no finding does not establish that every generated function and every business rule has been tested.
If Hackaiz prepares a code fix, it is a draft GitHub pull request that a human must review and approve. Nothing is merged automatically. Coordinate that review with the workflow you use to synchronize changes back to the builder, and test the affected behavior before release. Optional monitoring applies only to connected systems. Neither generating the app with AI nor connecting it to Hackaiz removes your responsibility for deciding which changes reach production.
Frequently asked questions
Take the next step
Was this article helpful?