Borrowed code, owned risk

Website Security Services for the Code You Did Not Write: Audit It, Fence It, Watch It

On November 16, 2026, a testing tool used by some of the largest brands on the web spent a day delivering an attacker's code to their visitors. Your site probably runs a dozen tools built the same way. This section is about knowing which ones, and deciding which ones stay.

November 19, 2026 · 11 min read

acmeoutfitters.com · checkout script scan
Everything the checkout loaded, desktop and iPhone
Nine scripts at checkout, three nobody can explain
Sample
  • PassCard entry in the processor's hosted fields
  • PassAnalytics tag has a named ownermarketing
  • Failcdn.acme-chatwidget.exampleadded 2021, no owner
  • WarnTesting tool loader on every pageno tests running
  • WarnTag manager publish rights11 people, 3 gone
  • Failpixel.acme-adnetwork.exampleagency gone
  • FailContent Security Policy sentnone
  • FailAlert when a script changesnot set up
Illustrative: sample script scan of a fictional outfitter's checkout

The short answer

Website security for third-party scripts comes down to three jobs: an audit that lists every script your pages load, who owns it and what it can reach; a Content Security Policy that tells browsers to refuse code you did not approve; and monitoring that flags when a script changes. DataCram responds the same business day, and each service starts from $575, depending on complexity, with a written estimate first.

On this page
  1. A trusted vendor, a sign-up form, and 178 organizations
  2. Your website is mostly other people's code
  3. Match the worry to the service
  4. Know it, fence it, watch it: the order that saves money
  5. Where the skimmer could not reach
  6. From $575, in writing, with a reply the same business day
  7. Five owners who should keep their money

A trusted vendor, a sign-up form, and 178 organizations

The attack that prompted this section never touched a victim's server. Optimizely's incident notice, posted on its status page on November 17, 2026, said an unauthorized party "exploited a vulnerability in Experimentation" and used it to add malicious JavaScript to some customer projects. Optimizely says it fixed the flaw, removed the injected code, secured the affected accounts and added protections.

The scale came from outside. Report URI, a company that sells script monitoring, reported on November 16, 2026, in a post it updated the next day, that it fetched 1,611 distinct Optimizely scripts and found 188 of them, belonging to 178 organizations, carrying the attacker's code. It says the real number is likely higher. Our account of the Optimizely hack covers the timeline and the cleanup for Optimizely customers.

This hub is about the lesson for everyone else, which is quieter and more useful. Those sites were not broken into. They were delivered to. The code came through a door their owners had opened on purpose, for a good reason, and then stopped thinking about. Almost every business website has doors like that.

Your website is mostly other people's code

A third-party script is a piece of code your page tells each visitor's browser to fetch from somebody else's server: the analytics tag, the chat bubble, the review stars, the testing tool, the heat map, the cookie banner. Once it arrives, the browser treats it as part of your page. Unless something tells the browser otherwise, it can read what a visitor types into your forms, change what the screen shows and send data anywhere on the internet.

That is the bargain. You chose the vendor. The vendor chose its own staff, its own suppliers and its own sign-up form. When any link in that chain fails, a stranger's code reaches your visitors wearing the vendor's badge. Security people call this a supply chain attack, which is a formal way of saying the delivery truck was trusted and nobody opened the box. Take a fair look at a typical small business site and you tend to find some version of this cast:

  • The chat widget from 2021, still loading on every page, connected to an inbox nobody has opened since the agency left.
  • A tag manager anyone who ever touched the account can still publish to, including the intern from three summers ago.
  • A testing tool that has run no tests since spring and still loads before your headline does.
  • A cookie banner that loads four scripts before it asks permission to load scripts.
  • A browser policy that says, in effect, trust everything this vendor sends, written long before the day the vendor gets hacked.
Goofy flat sticker of a burly longhorn bouncer at a velvet rope, squinting at tiny paper scrolls and stopping a masked bug, agave and sunset hills behind
Every script on your page got past the rope once; the question is whether anyone still checks.

None of these is a scandal. Each one solved a real problem on the day it went in. The trouble is that a script is not a purchase you make once. It is a standing invitation, renewed every time a visitor loads the page, to whoever controls the file at that moment.

Match the worry to the service

The eight pages in this section each sell one job. Find the sentence in the left column that sounds like your week, and start there. If two match, start with the one that involves payment cards or customer accounts.

What keeps owners up at night, and the service built for it
What you are thinkingThe serviceWhat you end up with
I cannot say what runs on our pagesThird-party script auditEvery script listed with its owner, its reach and a verdict
Any vendor could push anything to our visitorsContent Security Policy setupA browser rule set, proven in report-only mode before it enforces
Our processor asked about PCI 6.4.3 and 11.6.1Payment page script controlsAn evidence pack for your assessor; we do not certify compliance
Nobody would notice if a script changed tonightScript change monitoringAlerts when a script's contents, size or destinations change
We found code on the site that we did not writeTampered script responseEvidence kept, damage contained, the way in closed
Our testing tool loads a vendor's code on every pageServer-side testingTests decided on the server, with no testing script in the browser
The tag manager has become a junk drawerTag manager cleanupDead tags gone, a named owner for each survivor, fewer publishers
We run twenty plugins and keep addingReplacing risky pluginsCustom code where a plugin is not worth the exposure

If what you need is a cleanup after a classic break-in, the kind with spam pages and a red browser warning, that work lives on our malware removal and website security page. This section is about the code your site invites in through the front door, which is a different problem with a different fix.

Know it, fence it, watch it: the order that saves money

Owners who read about Optimizely tend to want the strongest lock first. That instinct costs money. A browser policy written before anyone has listed the scripts is a guess, and a guess either blocks the checkout or allows everything. Here is the sequence we follow, and the one we would suggest even if you never hire us.

  1. 1

    Make the list

    Load the important pages in real browsers and write down every script that runs, where it comes from and who wanted it. The script audit is this step done thoroughly, on desktop and on an iPhone.

  2. 2

    Cut what nobody will defend

    A script without an owner is a liability with a loading time. Removing it is the cheapest security work there is, and the page gets faster. Security and speed collide in other places too, as when a host turns off HTTP/2 as a precaution and nobody turns it back on. The tag manager cleanup is often where most of the cuts happen.

  3. 3

    Move the risky work off the page

    Some jobs do not need to run in the visitor's browser at all. Experiments can be decided on your server, and card numbers can be typed into your processor's hosted fields instead of your own form.

  4. 4

    Write the policy, then let it listen

    The Content Security Policy starts in report-only mode, where the browser reports what it would have blocked without blocking anything. Once the reports are quiet, it enforces.

  5. 5

    Watch for changes

    Scripts change on their own schedule. Monitoring checks what visitors actually receive and raises a flag when a file grows, shrinks or starts calling a new address.

  6. 6

    Decide who answers the phone

    Write down who you call if a script turns hostile, and who can reach your processor and your attorney. The response page describes the first hours.

Where the skimmer could not reach

One detail from November 16 deserves more attention than it got. The same Report URI analysis, published November 16, 2026, found that the payload skimmed card numbers, expiration dates and security codes from forms on a site's own pages, starting at 13:43 UTC. Cards typed into a payment company's hosted fields, the small secure boxes supplied by Stripe or Adyen, were out of its reach.

Read that as a design principle. The strongest defense on the day was not a clever scanner. It was a page that never held the thing worth stealing. The card box belonged to the processor, so a stranger's script on the merchant's page had nothing to grab. Our ecommerce development work starts from that assumption, and the payment page scripts service explains how the card industry's rules treat each kind of checkout.

The cheapest defense is knowing what runs on your pages. The strongest is a site that runs only what you chose.

The same thinking scales up. A site built for you can load the scripts you picked and nothing else, run its experiments on the server and enforce a strict policy from launch, because there is nothing to make room for. That is the case for custom websites, and for owners already on WordPress, for a planned move such as WordPress to Next.js. Our Athena versus WordPress comparison covers the content side. Custom code is not immune to bugs. What it gives you is the final say.

From $575, in writing, with a reply the same business day

Every service in this section starts from $575, depending on complexity, and you get a written estimate before work starts. Complexity here means real things: how many page types, whether there are logins and a checkout, how many vendors and tag manager containers, and what platform the site runs on. You approve the number, then we begin.

We respond the same business day; our hours are on the contact page. We do not promise nights, weekends or a finish date before we have seen the site. Ongoing script monitoring is added to a hosting and maintenance plan, which starts at $55 a month and covers managed hosting, updates checked on a staging copy first, daily off-server backups that are test-restored, security monitoring and uptime checks, with no long contract. The monitoring itself is quoted with your estimate.

What we will put in writing

  • Which scripts we found, who owns each and what it can reach
  • What we changed, and what we recommend you change
  • The policy we deployed, and every exception in it with a reason
  • What still carries risk after the work, stated plainly

What we will not put in writing

  • That your site cannot be hacked
  • That a policy blocks every possible attack
  • That you will pass a PCI assessment; we are not a Qualified Security Assessor
  • Legal advice about notifying customers after an incident

Not ready to spend anything? The free website health check and site audit flag the obvious problems, and is my site down answers the more urgent question. DataCram is a web studio in San Antonio, Texas, building for the web since 1999 and working with businesses across the US by phone, video and a shared project page. Read about the studio and see our published work before you hand anyone a login.

Five owners who should keep their money

A security vendor that finds danger everywhere is selling fear by the pound. Plenty of sites need less than this section offers, and a few need someone else first. If you see yourself below, take the advice and skip the invoice.

The five-page brochure site

Analytics, a contact form, no logins, no checkout, and you can name every script from memory. Delete what you do not use and try the ten-minute check on our script audit page. That may be the whole job.

The team that already keeps a list

If your developer maintains a script inventory and an enforced policy, and can show you both, keep them. A second opinion is worth having before a big change, not as a monthly habit.

The site that is hacked right now

An active break-in needs triage before tidying. Start with our hacked website emergency guide and the emergency repair hub. If customer data may be involved, call your attorney, your insurer and your payment processor first; our data breach first-hours page sets out the order.

The owner who wants a certificate

Nobody honest sells a document saying a website is safe. If that is what the board wants, we are the wrong call, and so is anyone who says yes.

The business with no technical owner

If no one inside will ever read an alert, monitoring is noise. A fractional CTO who owns the list and the decisions may be the better first hire.

Borrowed code is not limited to your website. The same trust problem rides along in the add-ons your staff install in their browsers, which our browser extension security page covers, and in the plugin layer of any WordPress site, covered in our plugin security notes. If your testing program is the main source of outside scripts, talk to whoever runs your conversion rate optimization before you cancel anything; the experiments can usually move rather than stop.

Questions owners ask about website security

It is an attack that reaches your visitors through a vendor you trust instead of through your own server. The attacker compromises a tool your pages load, such as a testing script or a chat widget, and the browser runs the tampered code because it came from an address you approved. The November 16, 2026 Optimizely incident is a large recent example; our post about it explains how it worked.

No, and be wary of anyone who says yes. What we can do is shrink what a stranger's code is able to reach: fewer scripts, a browser policy that refuses the ones you did not approve, card entry handled by your processor's hosted fields, and alerts when something changes. Each step lowers the odds and the damage. None of them is a promise.

If it takes payments, has customer logins or runs marketing tags you did not write, a policy is worth having. If it is a short brochure site with one analytics tag, cutting unused scripts may matter more. Either way, start with a list of what runs, because the policy is built from that list. Our CSP setup page explains the process.

Each one starts from $575, depending on complexity, with a written estimate before work starts. The number moves with the number of page types, logins, checkout, vendors and platform. Ongoing monitoring is added to a hosting and maintenance plan, which starts at $55 a month, and the monitoring is quoted with your estimate.

Usually not. Most tools can stay once each has an owner, a reason and a place in the policy. Some can move: experiments can be decided on the server, and card entry can move into your processor's hosted fields. The tools that go are the ones nobody can explain, and those are rarely missed.

We cannot promise that, and nobody honest in our position would. We are not a PCI Qualified Security Assessor and we do not certify compliance. We can build the script inventory, authorization records and change alerts that requirements 6.4.3 and 11.6.1 ask about, and hand them to you for your acquirer or assessor. Confirm your own scope with them.

Keep reading

Free · No obligation

Find out what your pages are running

Send your address and the worry that brought you here. We respond the same business day with what we would check first, and a written estimate before any work starts.

Which of these sounds like you? (pick any)

Rather talk now? Call (210) 346-0848.

CallGet a quote