Insights / Security
Seven security holes in quickly built apps
By Bakhrom Rikhsiboev, founder of First Door Studio. Last reviewed November 2026.
AI coding tools and fast freelancers have made it cheap to build a booking app, a client portal or an online store. That's good news. The catch is that these tools build what you ask for, not what you forgot to ask for. The same handful of gaps shows up again and again, and finding them doesn't take hacking skills. Anyone can open a browser's developer tools and read everything a site sends them.
Here are the seven we look for first, in plain terms.
- Secret keys sitting in the browser. If your app talks to an AI service or a payment provider, it uses a secret key. When that key is in the website's code, anyone can copy it and use it, and you pay for every request. The fix: keep secret keys on the server, and have the browser ask your server to make the call.
- Secrets saved in Git. Code is usually stored in a Git repository, often on GitHub. A key committed there, even once and then deleted, stays in the history. Automated bots scan public code for keys constantly. The fix: remove it from the history, then replace the key, because you have to assume it was already copied. Add a scanner that blocks keys from being committed.
- A database anyone can read. Popular app databases give the browser a public key. That's fine only when row-level security is on, which means rules on every table that decide who can see which rows. With it off, the public key reads every customer, email and order. Passwords are usually scrambled, but a full email list is enough to send convincing scam emails to all your customers. The fix: turn on row-level security for every table and test it with a regular user account.
- Records with no owner check. You open your invoice at an address ending in 1041. You change it to 1042 and see another customer's name, address and order. The fix: on every request, the server checks that the record belongs to the person asking.
- An admin page without a lock. Type /admin and you're in: delete users, issue refunds, export everything. Hiding a link isn't protection. The fix: admin pages require a login and an admin role, checked on the server for every action.
- Unlimited sign-ups. A script creates thousands of fake accounts overnight. If each one gets free AI credits, trial emails or text messages, the bill arrives in the morning, and the dashboard looks like you went viral. The fix: rate limits, bot protection, email verification, and a hard spending cap on anything billed per use.
- Payment confirmations that aren't checked. Payment providers tell your app "payment succeeded" through a webhook. If the app doesn't verify the provider's signature, anyone can send that message themselves and get a paid plan for free. The fix: verify the signature on every webhook, and confirm the payment with the provider before granting anything.
Check again after every change
These aren't one-time fixes. Each time code changes, especially when an AI tool rewrites part of it, one of these gaps can quietly reopen. Automatic security checks that run on every change, and a quick review before each release, catch them early.
Our launch checkpoints include the full list of 20 security items, from encrypted data and secure cookies to file upload limits and security headers.
This is general guidance, not a security audit or a penetration test. If you handle health, financial or card data, you may have extra legal requirements.
Want someone to look at your app? That's the app and website security review.