Prompt library

Fix prompts you can paste

Copy-paste prompts that fix real security problems in AI-built apps. Paste one into Cursor, Claude Code or your builder and it does the work — including the traps to avoid.

Why a prompt instead of instructions

If you built your app by describing it to an AI, the fastest way to fix it is to describe the fix the same way. You do not need to understand Row Level Security to paste a prompt that turns it on correctly — but you do need the prompt to be written by someone who does.

That is the part these are for. Each one names the specific trap the AI would otherwise fall into, because the quickest way to make a security error go away is usually to remove the protection that produced it. Ask an assistant to fix "new row violates row-level security policy" without constraints and there is a fair chance it disables the policy. The prompt below that error forbids exactly that, in writing, before it forbids anything else.

How to use one

  1. 01Pick the prompt matching your problem. If you are not sure which, run the free check on the homepage — it names the problem for you.
  2. 02Copy it whole. The constraints at the bottom are load-bearing; a trimmed prompt is how you get a fix that silences the symptom.
  3. 03Paste it into Cursor, Claude Code, Lovable, Windsurf or whatever you build with.
  4. 04Read what it says it changed, then verify from outside using the check on the same page. An AI reporting success is not evidence.

Database access

The largest single cause of leaked user data in AI-built apps. The browser talks to the database directly, so the database is the only thing that can decide who reads what.

Secrets and keys

AI tools put the key where the code that needs it lives. When that code runs in the browser, so does the key — and deployed JavaScript is scraped continuously.

Who is allowed in

Hiding a page from the menu is not access control, and being signed in is not the same as being allowed. These move enforcement to the server, where the visitor cannot edit it.

Sessions and abuse

Where the session lives decides how bad one injected script gets, and whether anything stops the same request arriving ten thousand times.

Browser-level protection

The cheapest protection available: no application changes, and it turns several classes of bug from serious into harmless.

Where to go next

Not sure which one you need?

Run the free check on your live site. It names the problems it finds and hands you the matching prompt, already filled in. About 30 seconds, no signup.

Check my site — free