#053Tutorial3 min
Audit an app built with Replit
Keep the running Replit deployment, source repository, and proof of control aligned.
Hackaiz team · updated 21/11/2026

01Start from the running deployment
A Replit project can have code in an editor while users reach a separately published version. For a useful security review, first identify the address your users actually visit and the source version intended to serve it. Include a custom domain if that is the production entry point. Do not substitute a temporary preview merely because it is easy to open: results for that address describe that environment, not every deployment created from the same project. Keep development and production clearly named in your own notes.
Hackaiz can use a website address and can inspect an authorized GitHub repository. On Connect, the connected-system route lets you connect GitHub, then choose a repository to scan. If you have not made the relevant Replit source available in an authorized repository, an address alone does not provide that source. This article does not describe a native Replit workspace login or a remote terminal inside your project. Select only the concrete resource you control and are authorized to test.
02Make the ownership proof reachable
Before any website or API analysis, establish domain control. You can add the DNS TXT record _hackaiz, serve the HTTPS file /.well-known/hackaiz-verification.txt, or give Hackaiz’s instruction to the AI that edits the project so it creates the file. The Replit platform entry lists public/.well-known/hackaiz-verification.txt as the project-side path. The file must contain the displayed verification value and must be present in the deployed application, not just in an unpublished working directory.
After publishing, check that the selected hostname is the one serving the proof. A custom domain and a development address are different targets. If the app serves a normal page where the text file should be, correct the publication or choose DNS if you control it. Do not proceed as though a verification failure were an ordinary warning in the security report. Proof comes before analysis. Review the chosen checks before launch: analysis work consumes credits, while unticked checks do not execute and are not billed.
03Examine secrets, queries, and AI actions
In repository findings, pay attention to where privileged work happens. A provider key read through a public browser variable can be copied by a visitor even when it was initially entered as configuration. Hackaiz includes rules for exposed AI provider keys, variables written into credential logs, SQL assembled from variable text, and command execution built from variables. An agent endpoint that forwards model output to a shell deserves a different repair from an ordinary visual bug. Supported detection patterns are starting points for review, not proof that every Replit app contains these weaknesses.
Use the finding’s evidence to make a narrowly defined change in your normal development process. Keep secrets out of copied prompts and test both permitted and forbidden requests. Customer-facing levels are Danger, To fix, and To watch, not a promise that the whole app has been assessed exhaustively. When Hackaiz prepares a GitHub repair, publication requires approval and produces a draft pull request for human review; it is never merged automatically. After release, compare the original scenario with the actual deployment. Correction verification requires replay, passing automated tests, and a matching corrected commit, not merely a new scan with fewer findings.
Frequently asked questions
Take the next step
Was this article helpful?