Skip to main content
Third + Grove — home
Perspective - Article

Ask Your Agency for Standing Authority to Patch

[ Author ]Justin Emond
[ Date ]11.19.26
[ Time ]6 Min Read

Before Black Friday, every brand should decide whether to give its agency two permissions.

One is pre-approval to apply an urgent security patch on the agency's judgment. The other is authority to take a site, feature, or app offline while it's being exploited. They're separate decisions. You can grant none, one, or both, but you should make the call now, in writing, instead of in the middle of an incident on the busiest weekend of the year.

Here's why we're asking, and how to scope each one.

Why Is This Coming Up Now?

AI models are now reading through years of code that nobody ever checked with AI. That means a lot more vulnerabilities are being found, all at once. As I've been putting it to our team and clients, we're in this interim period where all of this code has never been checked with AI before. So there's just a backlog of vulnerabilities they're finding.

The clearest example landed on October 21. PHP, the language behind WordPress, Drupal, Magento, and a lot of custom middleware, shipped fixes for two critical vulnerabilities that allow remote code execution without logging in. OpenAI said one of them was found by a research team using its model, and wrote that it expects "similar findings to increase sharply in the coming months." Five days after the fix shipped, CISA added both to its catalog of known exploited vulnerabilities and told organizations that couldn't update right away to take those systems offline until they could.

Shopify runs and patches its own platform, so your store's core isn't the exposure there. But most of the brands we work with run something next to Shopify: a content site, a blog, an integration layer, a custom app. That's where a finding like this lands.

My best guess is that this lasts months, then comes down, but not back to where it was.

What Happened to Shopify Stores in October?

Between October 10 and 19, merchants and researchers reported malicious code delivered through at least three third-party Shopify apps. When a shopper clicked checkout, the code sent them to a lookalike checkout page that collected their card details. Then it forwarded them to the real checkout, so the order still went through and nothing looked wrong. iTnews covered the incident, including the app developers' own statements. Shopify's checkout itself wasn't breached.

Two details matter for how you plan. First, the code didn't live in the stores' theme files. It arrived through settings the apps loaded from their vendors' servers, so a review of your own code wouldn't have found it. Second, at least one affected retailer stayed closed for days on Shopify's advice. Taking a store offline isn't a hypothetical. It happened last month.

Shopify has since added protection against unexpected redirects, and that helps. It's a platform-wide safety net, not a response plan for your store.

Why Is the Approval Email Now the Slow Part?

The usual loop goes like this. Your agency finds the issue, emails you, waits for a reply, and schedules the fix. On a normal week that takes a few days, and for years that was fine, because writing and testing the fix took a few days too.

That math has changed on both sides. Vulnerabilities are being found and exploited faster. And with agentic development, producing and testing a patch takes hours, not days. When the fix is fast and cheap, permission is the slow part. The PHP timeline makes the point: five days from patch to active exploitation. An approval email that sits over a weekend can use up most of that window.

Permission 1: Apply an Urgent Security Patch Without Waiting

This is pre-approval for your agency to apply a security fix on its own judgment, then tell you right away.

What it should cover: security patches rated critical or high by the vendor or a public advisory, updates to platforms and dependencies your site runs on, and disabling or updating an app that's the source of a known problem.

What it shouldn't cover: new features, design changes, or anything that changes how your store looks or sells. If a fix would change the customer experience, that's still your call.

The guardrails that make it safe:

  • A written severity threshold, so "urgent" means the same thing to both sides.
  • A named lead developer who owns the change.
  • The same review and QA pipeline as every other change. We walked through those gates in Is AI-Written Code Safe to Ship to Your Store?
  • A note to you as soon as it ships: what changed, why, and how to roll it back.

Permission 2: Take Something Offline While It's Being Exploited

This is the harder one, because it costs revenue in the moment. It's the authority to disable a feature, pull an app, put a content site into maintenance mode, or take the storefront down while an active exploit is running.

The trade is a few hours of lost sales against customers typing card numbers into a fake form. Most brands, asked in a calm moment, pick the first. The point of granting it now is that you won't be asked in a calm moment.

Scope it with three answers:

  1. Who on your side gets called first, and who the backup is when they don't pick up.
  2. What the agency can take down on its own, and what still needs a live yes. A compromised app is different from the whole storefront.
  3. What comes back first, and who signs off before it does.

How Do You Decide: None, One, or Both?

A few factors decide it:

  • How much third-party code runs on your storefront. Every app that loads a script is surface area you don't control.
  • What you run next to Shopify. A PHP-based content site or custom integration raises the stakes on Permission 1.
  • Who's reachable over peak. If nobody on your team answers at 2 a.m. on Black Friday, an approval process is really a waiting process.
  • Your appetite for risk. Some brands would rather lose an hour of sales than take any chance with customer data. Some wouldn't.

Here's how we'd apply it. If you have a large app footprint and a lean internal team, grant both, with tight written scope. If you have your own engineers on call around the clock through peak, keep Permission 2 in-house but grant Permission 1. "None" is a valid answer only if someone on your side has the authority and the availability to act within hours.

"We Don't Want an Agency Making Calls on Our Site"

That's a reasonable instinct, and we'd be wary of any partner who brushed it off. Standing authority doesn't mean unchecked authority. The scope is written down and you can revoke it anytime. A named person is accountable for every change. Every change goes through the same review, testing, and QA it always would. And nothing happens without you hearing about it the same day.

What you give up is the delay, not the oversight.

What to Do Before Black Friday

  1. Decide on each permission separately, and confirm it in writing. An email reply is enough.
  2. Define "urgent" with a severity threshold you both agree on.
  3. Name your contacts for a takedown decision, with a backup and real phone numbers for peak weekend.
  4. Review which apps load code on your storefront, and remove the ones you no longer use. Shopify now shows app security details under Settings > Apps.
  5. Agree on how you'll hear about it after a patch or takedown, and how fast.

None of this needs a code change, so it fits inside a code freeze.

This is now a different threat environment, and we have to operate differently. Making the decision before peak is how you stay in control of it.

Reduced motion disabled