What the browser extension risk agent does
Your workforce installs browser extensions freely, and those extensions can read page content, cookies, and credentials on every site someone visits. They also update silently, so an extension that was safe last month can turn malicious after an ownership transfer or a supply-chain compromise, without anyone clicking anything.
The browser extension risk agent reviews every extension installed in your environment: its requested permissions, its web store listing, and what changed in the latest version. It assigns each one a verdict and brings the ones that need a human decision to you. When you mark an extension not permitted, Nudge Security can enforce that decision everywhere: it disables the extension across your organization, tells each affected person why and asks them to uninstall it, and blocks future installation attempts.
Analysis and enforcement both depend on the Nudge Security browser extension. Browsers without it are outside the agent's reach—there's nothing to analyze and nothing to enforce. See Deploy the browser extension.
Find the agent
Go to Automations > Agents. The Browser extension risk agent card shows the agent's status and an on/off toggle:
Select View all activity to open the agent's page, where you can see everything it has analyzed.
Select the settings icon to change how the agent acts on its verdicts. You can also select Configure on the agent's page.
The agent's page also shows its Status and an Enabled toggle. Turn the toggle off to stop the agent from analyzing new extensions.
How analysis gets triggered
While the agent is enabled, it runs automatically whenever a new extension, or a new version of one you already have, shows up on a monitored browser:
Version check. If this exact version on this browser product has already run for your organization, it doesn't run again. If another customer's browser has already put it through analysis, you get the stored verdict immediately.
Metadata fetch. Nudge pulls the listing from the web store: name, publisher, permissions, version, and ratings.
Triage. If an earlier version was analyzed, the agent compares the two and re-analyzes only when the changes warrant it—new permissions, new remote endpoints, or a changed publisher. If nothing meaningful changed, the previous verdict carries forward. If the extension has never been analyzed, triage criteria decide whether a full analysis is needed.
AI analysis. The agent inspects the extension package and its metadata and produces a verdict plus a plain-language explanation you can share with employees.
Routing. The agent records the verdict against the version and acts on it based on the verdict and your settings.
Each version is analyzed once per browser and the result is shared across every customer, so most discoveries resolve in seconds. A version nobody has seen before typically has a verdict within 15 minutes.
Verdicts are per browser, because each browser extension store ships a different package. The same extension on Chrome and Edge is two records, analyzed separately, with two approval statuses and two decisions to make. When you block one, Nudge shows you the sibling records so it's clear what the decision hasn't covered.
The four verdicts
Normal. Permissions and publisher match what the extension claims to do. The agent permits it and the workflow ends. The activity record shows an outcome of Permitted.
Risky. Broad or unnecessary permissions, an unverified publisher, or a version change that expanded access. With Review risky extensions turned on, the analysis comes to you for a decision. Turn that setting off and the agent permits risky extensions, which keeps your review queue focused on malicious and compromised verdicts.
Malicious. The extension does something it shouldn't and its listing doesn't disclose it—exfiltrating data, injecting code, or hiding its behavior. The analysis always comes to you for a decision.
Compromised. A once-legitimate extension that changed hands or shipped a malicious update. The analysis always comes to you for a decision.
A failed or malformed analysis never defaults to Normal. Nudge retries it, then escalates it to Nudge Security engineering, and the version stays visible to you as unassessed so you can see the gap.
Configure the agent
Select Configure on the agent's page to open its settings, then select Save when you're done. Settings changes apply only to extensions discovered after you save.
Observe mode (monitor only). The agent analyzes every newly discovered extension and records the verdict, but takes no action. Use it to see how the agent judges your own environment before you let it act.
Review risky extensions. Sends extensions with a Risky verdict for human review before anything happens. Turn it off and the agent permits them.
Nudge the following user to review agent analysis. Choose the person who reviews the agent's work. Every analysis that needs a human decision goes to them as a nudge with the verdict, the permissions requested, and the agent's reasoning, and they decide whether to permit the extension.
Automatically disable not permitted extensions. When an extension's approval status is set to Not permitted, Nudge disables it across your organization and prevents it from being re-enabled or installed elsewhere. See What happens after you mark an extension not permitted below.
Disable compromised & malicious extensions right away. The agent disables the extension the moment the analysis lands, before anyone reviews it, so exposure is measured in minutes rather than days. The review still happens, and if you permit the extension, end users can re-enable it.
New to the agent? Start with observe mode on. Review a week or two of verdicts on your own extensions, then turn it off and choose how much you want the agent to act on its own.
Review a flagged extension
Malicious and Compromised verdicts, and Risky verdicts when Review risky extensions is on, need a human decision. The reviewer you chose in the agent's settings gets a nudge, and you can also review from the agent's page in Nudge Security. Either way, you make one of two decisions.
Permit the extension
The extension stays available and the activity record shows an outcome of Permitted. The agent's verdict stays on the record, so an extension can show a Malicious verdict with a Permitted outcome—that's a documented human decision, not a contradiction.
When you permit an extension the agent flagged, add a clarification explaining why. The reason matters:
The verdict is wrong. This is real feedback on the analysis, and it's what improves verdict quality over time.
The verdict is right and we accept the risk. A business decision, not a correction. It's never counted against the agent's accuracy.
Mark the extension not permitted
This sets the extension's approval status to Not permitted, and the activity record shows an outcome of Not Permitted. If Automatically disable not permitted extensions is on, enforcement starts right away.
You can't save Not permitted without an employee explanation of why. For any version the agent analyzed, the explanation is already written and pre-filled—accept it or edit it. For a version triage skipped, you write it yourself. That text is what your people see in the nudge, so nobody is ever told to uninstall something without a reason.
What happens after you mark an extension not permitted
With Automatically disable not permitted extensions on, the Nudge Security browser extension enforces your decision. It keeps your organization's list of not-permitted extensions locally and acts on that list directly, with no server round trip:
Every instance is disabled across your organization on that browser product—every user, every browser of that product, no matter where the flagged version was found.
Each affected person is nudged with the extension name, the browser product it applies to, the reason it isn't permitted, and an uninstall button.
Re-enabling doesn't work. If someone switches the extension back on, it's disabled again immediately and they're nudged again.
Installation attempts are blocked at the point of install or prevented at the marketplace, and the person is told why. The block covers every version of that extension on that browser product, including versions never analyzed and versions released after your decision.
A browser seen for the first time after your decision gets the same treatment as soon as it has the list.
Uninstalls have three outcomes, each reported separately: uninstalled, declined where the person dismissed the browser's confirmation, and unavailable where the extension was installed by policy and can't be removed this way. Declined and unavailable both leave the request open. Unavailable falls back to manual instructions, since no button will ever resolve it.
Nudge disables extensions automatically but never uninstalls them silently. The browser shows its own confirmation and the person can decline. Approval status changes reach browsers on the next sync—until then, the previous state applies in both directions.
Nudge templates
Two templates carry these messages, and both ship with default copy you can edit for your organization:
Blocked installation—shown when someone tries to install a not-permitted extension.
Disable and remove—shown when an installed extension is disabled. This one carries the uninstall button, with manual instructions as a fallback.
Each nudge names the extension, the browser product, and your reason for blocking it. Naming the browser product isn't cosmetic: someone running the same extension on two browsers needs to know which copy to remove. If a person has several outstanding extension nudges, they land in one tab rather than one tab each, and resolving one entry leaves the others alone.
Track what the agent is doing
The agent's page lists its Recent activity: every extension the agent has analyzed, with its Verdict, Outcome, and Date. Filter by verdict, outcome, or date, or use Search analysis to find a specific extension. The same extension on two browser products always appears as two rows, so they're never mistaken for duplicates.
Select a row to open its details. The panel shows the item's Status, Outcome, and Verdict, a short summary of the agent's analysis, and a Steps timeline showing what happened, oldest first:
Trigger—the install the agent detected.
Assessed risk and Recommendation—the agent's risk rating and the action it recommends, such as Remove, along with how many signals it reviewed.
Nudge to the reviewer—when the item needed a human decision.
Decision—who decided, what they decided, and any clarification they gave.
Outcome—the final result, such as Permitted.
Use the arrows at the top of the panel to move to the previous or next item.
Read the full analysis
Select View full analysis in the details panel to open the complete report. You can also find it on the extension's page under Inventory > Browser extensions. The report contains:
Verdict—the risk level, with Why this verdict? explaining how the agent got there.
TL;DR—the answer to "should I care about this?" in two or three plain-language sentences you can paste into a ticket or share with employees.
Summary—a technical account of what the extension does, where it sends data, and how that compares to what the store listing claims.
Risk profile—how many findings landed at Critical, High, Medium, Low, and Info.
Findings—each one names a specific behavior and its severity. Expand a finding to see the evidence, or select Expand all.
Detailed technical summary—the full write-up for when you need to defend the decision.
You can also re-run the analysis against the latest version of the extension. At the bottom of the report, use Was this analysis helpful? to tell us how the agent did.
Reverse a decision
Only a person can reverse Not permitted. No verdict does it automatically, including a clean verdict on a newer version. To make that possible, Nudge keeps not-permitted extensions in your extensions list even after installs drop to zero, along with their approval status and most recent verdict. Extensions with any other status still drop off the list at zero installs, as they always have.
If the agent disabled an extension right away and you then permit it, end users can re-enable it.
Good to know
Approval status applies to the whole extension, not a single version. Blocking it blocks every version on that browser product.
Decisions don't cross browser products. Blocking an extension on Chrome doesn't stop someone from installing it from the Edge store—you need to block that record too.
The agent only analyzes versions it finds on your browsers. It doesn't backfill an extension's full version history.
Settings changes apply to newly discovered extensions. They don't revisit extensions the agent has already handled.