ScreenConnect, SCKit, and CVE-2026-84869: What to Check

TL;DR

The November 2026 ScreenConnect incidents involve two separate security weaknesses and several related attack behaviors. CVE-2026-127063 affects the independently developed RMM+ extension, whose default API secret could allow unauthorized remote commands. CVE-2026-84869 affects vulnerable ScreenConnect clients, allowing certain guest-initiated file transfers and execution actions without proper technician authorization. The client vulnerability was fixed in version 26.6.5, released September 8, 2026. ConnectWise had not announced a fixed RMM+ extension version as of November 18 and instructed affected partners to keep the extension disabled.

Huntress and Microsoft documented attackers using ScreenConnect management functionality to install rogue remote-access clients and a toolkit called SCKit. Huntress observed an initial November 11 wave involving at least 288 hosts across 48 organizations, followed by another wave involving 205 hosts across 41 organizations on November 13. SCKit was designed to reach technician workstations and potentially spread through their management sessions, but Huntress had not observed that particular propagation chain succeeding. A separate investigation confirmed exploitation of CVE-2026-84869 against a technician workstation.

There are three distinct defensive outcomes: block unsafe extension access, patch technician clients, and eradicate unauthorized remote access already established. Completing one does not prove the other two.

In This Article

Introduction

Consider a hypothetical managed service provider operating ScreenConnect to support several hundred customer devices.

The hosted management instance is current. The technicians use multifactor authentication. Endpoint agents report healthy, and normal remote sessions are working.

Then security identifies an unexpected ScreenConnect client installed on several customer machines.

The new client is not associated with the provider’s approved management instance. Some endpoints contain scripts and scheduled tasks associated with unauthorized remote access. The activity appears to have originated through the remote-management environment itself.

The immediate reaction is to update ScreenConnect.

That is necessary, but it does not answer every question.

The server might already be current while a vulnerable third-party extension remains installed. Technician workstations may still have older host clients. An unauthorized client installed during the intrusion can continue communicating with an attacker-controlled relay after the original access path is disabled.

If technicians reconnect to suspicious devices before the environment is understood, they may expose another privileged endpoint.

This is the operating problem exposed by the November 2026 ScreenConnect investigations.

The incident is particularly consequential for managed service providers because remote support infrastructure connects administrative identities to multiple customer environments. A single management instance can authorize actions across many endpoints, while a technician workstation may have access to additional management consoles, credentials, and business systems.

This article applies the series’ enterprise risk model to separate the vulnerabilities, documented malicious activity, and defensive decisions. It provides an assessment and remediation framework for ScreenConnect administrators, MSP security teams, and enterprise customers.

It does not treat SCKit’s apparent AI-generated code as proof that an autonomous AI framework controlled the campaign. It also does not merge the ScreenConnect incidents with the separate October ConnectWise Cloud Backup application-trust compromise.

The Three ScreenConnect Components That Matter

ScreenConnect, formerly ConnectWise Control, has three operational roles.

The management instance hosts the web console, manages sessions, and provides relay functionality. A technician uses a local host client to conduct interactive support. A managed endpoint runs an installed access client or connects as a guest during a support session.

These components form a powerful remote-administration workflow.

They also create different security boundaries.

The Management Instance Can Issue Remote Commands

An authorized technician or administrative workflow may use ScreenConnect to execute commands on a managed endpoint.

Microsoft’s November 13 investigation explains that the endpoint client receives commands sent through the ScreenConnect server. These commands do not require the technician’s workstation to execute or relay the command locally.

A compromised management workflow can therefore affect multiple managed endpoints even if no technician is actively connected to each machine.

That distinction explains the initial SCKit deployment activity.

The Technician Client Is a Separate Execution Boundary

When a technician joins a session, the local host client participates in file transfers and other session operations.

CVE-2026-84869 affects this client-side boundary.

Under vulnerable conditions, a guest can cause the technician’s client to accept and execute files without the expected authorization or confirmation.

The management server is not the vulnerable component for this CVE.

Updating the hosted server does not establish that every technician’s installed client is fixed.

Third-Party Extensions Add Their Own Authority

ScreenConnect supports extensions that add integrations and administrative functions.

The independently developed RMM+ extension family exposes an API for remote command execution. Its authentication is separate from the standard technician console sign-in.

That means the extension’s own access-control design matters even when ScreenConnect’s normal user accounts, roles, and multifactor authentication are configured correctly.

The November incident demonstrates why third-party extensions must be reviewed as privileged software, not simply as optional interface enhancements.

The Two CVEs Are Not Interchangeable

The most important technical distinction is that the November extension issue and the September client issue affect different components.

CVE-2026-127063CVE-2026-84869
Affected componentThird-party DataComm RMM+ extensionScreenConnect client
VulnerabilityDefault API secret can allow unauthorized extension operationsClient processes certain guest file-transfer and execution actions without proper authorization
Primary boundaryManagement extension API to managed endpointsGuest endpoint to connecting technician
CVSS score9.3 Critical, CVSS v4.09.9 Critical, CVSS v3.1
PublishedNovember 12, 2026September 8, 2026
ScopeRMM+ extension deploymentsScreenConnect clients before 26.6.5
Vendor remediationDisable the extension; no fixed RMM+ release announcedUpdate to 26.6.5 or later
Important limitationRemoving the extension does not remove previously deployed malwarePatching the client does not remove rogue clients or establish the original management entry path

The two scores use different CVSS versions and should not be treated as a direct numerical ranking of enterprise exposure.

A vulnerable technician client can represent a significant risk even if the RMM+ extension was never installed.

An exposed RMM+ extension can create a remote-command risk even when the installed ScreenConnect clients are current.

An already compromised endpoint can remain compromised after both software issues have been addressed.

CVE-2026-127063: The RMM+ Extension Authentication Failure

ConnectWise’s November 12 disclosure identifies a default API-secret weakness in the independently developed DataComm RMM+ extension.

The extension exposes functions that can send commands to managed agents. If the API credential remains at its unsafe default, unauthorized requests may be accepted.

Huntress identified a specific configuration problem in RMM+ versions 1.0.31 and 1.0.32.

In those builds, an installation with no access key configured could accept an empty access key. This allowed requests to reach the extension without an ordinary technician console sign-in.

Huntress correlated this behavior with incidents in which remote commands appeared through RMM+ while corresponding technician login activity was absent.

There is a source-scope distinction worth preserving.

Huntress specifically analyzed versions 1.0.31 and 1.0.32. ConnectWise’s broader CVE advisory does not limit its response guidance to those two builds and does not announce a fixed extension version.

As of November 18, the appropriate action is therefore not to assume that selecting a different RMM+ version or setting a stronger key completes the vendor’s required remediation.

ConnectWise instructs on-premises partners to disable the extension and not re-enable or reinstall it pending further guidance.

The vulnerability is specific to the third-party extension. It is not a general authentication vulnerability in ScreenConnect’s core server or client.

CVE-2026-84869: Unauthorized File Execution on Technician Clients

ConnectWise disclosed CVE-2026-84869 on September 8, 2026.

The flaw concerns file-transfer handling in ScreenConnect Support and Access sessions. Under affected conditions, a guest can initiate a transfer and execution action that the technician’s host client processes without the required authorization or confirmation.

The resulting action may execute with elevated privileges.

ConnectWise fixed the vulnerability in ScreenConnect 26.6.5 and later.

CISA added it to the Known Exploited Vulnerabilities catalog on September 11, 2026. Huntress later documented successful exploitation against technician workstations.

The affected execution path is particularly dangerous because the technician’s workstation is often an administrative hub. It may contain management consoles, browser sessions, approved support tools, and access to multiple customers.

The immediate patch requirement is to verify the installed version on each technician workstation. A current cloud console is not sufficient evidence.

ConnectWise also recommends keeping endpoint access agents updated.

Where client remediation cannot be completed immediately, its September bulletin provides an interim measure: remove the TransferFiles permission from affected roles and session groups. Older versions may label the permission TransferFilesInSession.

That is a temporary exposure reduction, not a substitute for patching.

What Happened During the November 2026 Campaigns

The timeline matters because the disclosures, attempted propagation, and successful exploitation occurred at different times.

DateFindingEvidence boundary
September 8ConnectWise released ScreenConnect 26.6.5 to address CVE-2026-84869.Fix for a client-side file-transfer vulnerability
September 11CISA added CVE-2026-84869 to its Known Exploited Vulnerabilities catalog.Exploitation was established, but not every later incident was caused by that CVE
November 11Huntress observed unauthorized ScreenConnect command activity affecting at least 288 hosts across 48 organizations.Rogue-client installation and SCKit-related deployment activity
November 12ConnectWise disclosed CVE-2026-127063 affecting the RMM+ extension.Separate extension authentication vulnerability
November 12Huntress observed successful execution of a guest-pushed malicious file on a technician workstation.Confirmed CVE-2026-84869 exploitation
November 13Huntress identified another SCKit-related wave affecting 205 hosts across 41 organizations.Further deployment activity; the full proposed worm propagation chain was not confirmed
November 16Huntress reported at least 688,000 guest-originated file-execution requests across 42 MSP ScreenConnect servers during its October 20 to November 16 review period.Requests, not 688,000 successful executions
November 16ConnectWise marked its ScreenConnect investigation complete while maintaining required protective actions.Vendor investigation status, not customer-by-customer security clearance

The two Huntress SCKit waves should not be converted into a single count of unique affected organizations without checking for overlap. They are distinct observations from Huntress’s monitored population.

Similarly, ConnectWise reported identifying approximately 400 partner environments with the relevant extensions installed and disabling those extensions in hosted environments as a precaution.

That is an extension-installation population, not a confirmed compromise count.

How SCKit Turned Remote Support Into an Attack Path

Huntress and Microsoft observed attackers issuing commands through ScreenConnect management instances and deploying unauthorized remote-access clients to managed endpoints.

The commands used the legitimate remote-execution capability of the management platform. On affected Windows endpoints, that activity could run with SYSTEM privileges.

The attacker then installed an additional ScreenConnect client configured to communicate with attacker-controlled infrastructure.

That second installation mattered because it established remote access independent of the legitimate MSP’s ScreenConnect instance.

The Observed Endpoint Deployment

Huntress described an installation chain that downloaded an attacker-controlled client installer and retrieved additional encrypted toolkit components.

The SCKit toolkit contained PowerShell scripts, batch scripts, command files, and configuration data.

Microsoft’s analysis identified a machine-specific payload retrieval process and components that established persistence through scheduled tasks and registry Run entries.

The toolkit also attempted to manipulate Microsoft Defender exclusions and interfere with normal ScreenConnect client operation.

The significant operating consequence is that the compromised endpoint might continue to appear manageable through familiar remote-support software while also communicating with an unauthorized relay.

A remote-access client’s executable name is therefore not sufficient to establish whether it is legitimate.

Administrators need to verify its installed instance, service configuration, associated relay, enrollment, and deployment provenance.

RelayTrap Was Designed to Reach Technicians

One SCKit component, RelayTrap, was designed to interfere with the legitimate endpoint-side ScreenConnect client.

The toolkit contained a scheduled task named TrapWatch, supporting files under C:ProgramDataRelayTrap, and a registry Run value named RelayTrapLiteAlert.

Huntress reported that its analyzed code positioned an attacker-controlled replacement client to interact with the normal technician connection workflow.

The intended next stage was to transfer an additional payload to a technician workstation when that technician connected to the affected endpoint.

From there, other components attempted to obtain the technician’s authenticated ScreenConnect browser session and use access to the management console.

The code included functionality intended to distribute further payloads through the ScreenConnect Shared Toolbox.

That architecture is significant because the technician’s permissions determine how many additional customer endpoints could become reachable if such a takeover succeeded.

The Intended Chain Was Not the Observed Outcome

Huntress explicitly distinguished the toolkit’s reconstructed capabilities from what it had observed during incident response.

It did not observe the SCKit endpoint-to-technician-to-panel propagation chain completing successfully. The attempted management-console takeover failed in the recovered host logs it reviewed.

The malware’s functionality and the campaign’s confirmed results must therefore remain separate.

The following table separates observed activity from reconstructed or unconfirmed propagation.

StageEvidence status in Huntress’s investigation
Unauthorized remote-access deploymentObserved in the investigated endpoint activity.
Endpoint-to-technician propagationReconstructed toolkit behavior; the full propagation chain was not observed completing.
Management-console takeoverThe attempt failed in the recovered host logs Huntress reviewed.
Broader customer impactRequires evidence of successful access and subsequent actions in each environment.

The defensive implication is still serious.

Even an unsuccessful propagation mechanism may leave unauthorized remote-access clients, persistence, modified configuration, and compromised endpoint trust.

The appropriate response is to investigate the observed compromise and the potential technician exposure, not to report the entire proposed propagation path as completed.

The Separately Confirmed Technician Exploitation

Huntress’s later investigation of CVE-2026-84869 established that real attackers were using the client-side vulnerability against MSP technicians.

In one November 12 incident, an attacker-controlled guest sent a malicious file that the technician’s vulnerable ScreenConnect host client executed.

The activity resulted in a new unauthorized ScreenConnect client connecting to an attacker-operated relay in approximately nine seconds.

Huntress observed the malicious file reporting to a Telegram bot, retrieving an installer, and starting the remote-access component.

On November 13, another technician client executed an attacker-controlled test file. Huntress assessed that event as an apparent canary used to verify the technique.

The researchers also found earlier October attempts against additional technicians, including attempts to place a script into a Windows Startup directory.

Those observations establish real exploitation of CVE-2026-84869.

They do not establish that the separately reconstructed SCKit console-takeover chain succeeded.

Why the File-Execution Requests Matter

Huntress identified at least 688,000 guest-originated “run file” requests across 42 MSP ScreenConnect servers from 29 source IP addresses between October 20 and November 16.

The attacker was repeatedly attempting file-execution actions, apparently waiting for a vulnerable technician to join an eligible session.

This is different from 688,000 successful executions.

A request might be made while no technician was connected. A client might already be patched. The session might not have the required conditions.

Huntress documented a near miss in which a request arrived shortly after a technician had disconnected. It classified that event as unsuccessful.

The distinction matters for incident severity and response scope.

A high-volume attempt against a server establishes attack pressure. A vulnerable host client connected during the relevant request establishes potential exposure. Process telemetry showing the file actually executed establishes a stronger finding.

For MSPs, this makes session correlation essential.

The Assessment Should Begin Without Reconnecting to Suspect Endpoints

There is an important safety constraint for this incident.

Do not open an interactive ScreenConnect session to a suspected SCKit-infected device simply to investigate it.

Huntress, Microsoft, and ConnectWise all advise using separate security tooling or a trusted out-of-band administration channel.

The reason is specific to the incident: connecting a technician can trigger the endpoint-to-technician interaction the malware was designed to exploit.

An incident team should first identify the affected management instances, endpoints, and technicians using available records and separate response capabilities.

If a technician has already connected to a suspected affected machine, include that workstation in the investigation rather than assuming the support session was harmless.

A clean workstation used for response should not depend on the potentially compromised ScreenConnect session or browser state.

A Practical ScreenConnect Exposure and Remediation Runbook

The following sequence is intended for authorized MSP and enterprise security teams. Product-specific steps must follow the applicable ConnectWise advisory and the organization’s incident procedures.

Establish the Administrative Scope

Inventory the ScreenConnect management instances in use, including hosted and on-premises installations.

For each instance, identify the accountable owner, version, installed extensions, administrative identities, enabled integrations, technician clients, endpoint access agents, and customer environments reached through that instance.

Do not treat a current server version as evidence that all clients or extensions are safe.

Determine whether the environment contains any of the five third-party extensions that ConnectWise instructed partners to disable:

  • Labtech Plus
  • RMM Plus
  • RMM+
  • RMM Plus Password Link
  • RMM+ Password Link

These extensions are independently developed components. The Labtech Plus extension is not synonymous with the ConnectWise Automate product.

Disable the Identified Extensions

ConnectWise reported that it disabled the affected extension family in cloud-hosted environments where it could take action.

On-premises partners must perform and verify the restriction themselves.

The November 16 ConnectWise advisory instructs partners to keep the five identified extensions disabled and not re-enable or reinstall them pending further guidance.

Do not substitute a strong RMM+ access key for this required vendor action.

The separate CVE-2026-127063 advisory states that no fixed RMM+ version had been announced.

If an extension is necessary for a legitimate business workflow, record the functional impact and establish an approved alternative rather than silently restoring the vulnerable integration.

The presence of an extension alone is not proof of compromise. Its exposure and activity still require separate investigation.

Verify Technician Client Versions

Enumerate the installed ScreenConnect host clients used by technicians, including temporary, branded, and older clients that may remain on administrative workstations.

Verify that every relevant technician client is running ScreenConnect 26.6.5 or later.

Use ConnectWise’s supported host-client reinstallation or update procedure. Keep endpoint access agents current as instructed by the vendor.

Review server connection history for previously used client versions, because the currently installed client may differ from the version used during a suspicious session.

During Huntress’s investigation, vulnerable technician clients from older 22.x, 24.2, and 25.8 release families remained in use weeks after the September fix.

A server inventory or cloud-console status does not resolve that gap.

Where the client cannot be updated immediately, evaluate the vendor’s temporary TransferFiles permission restriction across affected roles and session groups.

Document its expected operational impact and verify enforcement.

Preserve Console and Endpoint Evidence

Before clearing suspicious sessions, tools, extensions, or queued work, preserve the records needed to reconstruct the activity.

Relevant ScreenConnect evidence includes administrative and extension changes, remote command history, Toolbox operations, session participation, authentication records, and any available extension API activity.

Huntress documented session-database events that can help identify suspected CVE-2026-84869 attempts.

In the investigated ScreenConnect data, RanFiles corresponded to EventType 85, SentFiles to EventType 86, guest connections to ProcessType = 2, and host connections to ProcessType = 1.

Those values are source-specific investigative references, not a universal schema guarantee across all builds.

Where authorized, examine a preserved copy of the relevant session database rather than performing intrusive queries against a live production database during an incident.

A guest-originated file-execution event should be correlated with a technician host connection at the relevant time. That overlap helps establish exposure.

Then correlate the session with endpoint process execution, file creation, registry modification, and network activity to determine whether a payload actually ran.

Identify Unauthorized Endpoint Clients and Persistence

On suspected affected Windows endpoints, review the installed ScreenConnect clients and their configured relay destinations.

Look for unexplained additional ScreenConnect services, unexpected executable locations, installer artifacts, and command histories.

The SCKit investigations identified several relevant artifacts, including:

Investigative artifactWhat it may indicateImportant caveat
C:ProgramDataRelayTrapSCKit toolkit and trap-related filesDirectory presence needs validation against observed activity
TrapWatch scheduled taskRelayTrap watcher persistenceAbsence does not rule out other variants or removed artifacts
RelayTrapLiteAlert Run valueAutomatic startup associated with the toolkitReview the full registry context and originating process
sckit-decoder.ps1Decoder or payload-retrieval stageFilename may change and execution may have failed
Panel-Lateral.ps1Component intended for console or lateral activityFile presence does not prove successful lateral spread
Unexpected second ScreenConnect clientPotential unauthorized remote accessConfirm instance identity and relay before removal
Windows Update-like ScreenConnect namingPossible disguised remote client in reported activityLegitimate Windows Update components must not be assumed malicious by name alone
Unapproved Microsoft Defender exclusionsAttempt to weaken endpoint protectionAttribute changes to the responsible process or administrator

Huntress cautions that file hashes are relatively weak for this campaign because scripts and compiled components can vary between endpoints.

Use hashes and reported network indicators as enrichment, but prioritize behavior, executable provenance, process relationships, service configuration, and remote relay destinations.

A compromised endpoint can appear in an approved management inventory while also running an unauthorized client. Conversely, an unfamiliar client name should be investigated before it is deleted, particularly where customers legitimately use multiple remote-support providers.

Investigate Technician Workstations Independently

Review technician workstations that connected to affected or suspicious sessions.

Look for unexpected child processes launched by the ScreenConnect host client, particularly command shells, PowerShell, script interpreters, or installers originating from the client’s temporary working directories.

Huntress observed suspicious file execution, unexpected installer activity, and an unauthorized remote-access client disguised as Windows Update in its confirmed technician compromise.

Include browser session access and management credentials in the scope of investigation.

A technician’s workstation may contain privileged access to customer RMM instances, professional services automation systems, password vaults, identity platforms, and administrative portals.

If evidence supports compromise, isolate the workstation through trusted tooling and assess credentials and active sessions reachable from that device.

Do not assume that changing the ScreenConnect password terminates every existing browser or identity-provider session.

Remove Established Access and Validate Recovery

Once evidence is preserved, use the approved incident-response process to remove unauthorized clients, tasks, services, modified security settings, and compromised credentials.

Microsoft cautions against running cleanup code recovered from SCKit as a remediation shortcut. Its analysis found behavior in which cleanup could lead to redeployment of the toolkit.

Recovery should use trusted administration, known-good software sources, and appropriate rebuild procedures when endpoint integrity cannot be established.

Before reconnecting devices, examine pending ScreenConnect commands and Toolbox items that might execute when connectivity resumes.

Verify the approved ScreenConnect relay, the installed legitimate client, effective endpoint protection, and the absence of recreated persistence.

The service owner should accept the operational state, while security verifies that unauthorized authority remains restricted.

Match Each Detection to the Correct Attack Boundary

A security team can receive several ScreenConnect-related alerts without immediately knowing whether they represent extension abuse, unauthorized software installation, or a technician client exploit.

The distinction changes the investigation.

ObservationPrimary hypothesisCorroborating evidenceImportant limitation
Remote commands appear without expected technician login activityExtension API or another noninteractive management path was usedExtension configuration, API records, Run Command history, originating infrastructureDoes not independently prove exploitation of RMM+
Many endpoints install an additional ScreenConnect client within a short periodUnauthorized management-plane deploymentCommon command history, installer provenance, relay destinations, service creationDoes not prove every endpoint executed SCKit successfully
TrapWatch or RelayTrap files appearSCKit persistence or attempted propagationProcess timeline, task creation, registry changes, toolkit artifactsDoes not establish successful technician takeover
Guest session repeatedly requests file executionPossible CVE-2026-84869 exploitation attemptServer session records, client version, simultaneous technician connectionRequest count is not an execution count
ScreenConnect host client unexpectedly launches scriptsPotential technician-side executionHost-client process tree, file origin, command-line contextLegitimate tools can execute scripts; approval context matters
Unauthorized relay connection appearsPossible rogue remote-access clientExecutable path, service registration, client ID, destinationA single unfamiliar address requires investigation, not automatic attribution
Technician browser session is used unexpectedlyPotential session compromiseBrowser and identity-provider logs, management-console actions, session revocation stateRequires evidence of unauthorized use, not just a theoretical capability

This matrix is useful because each finding points to a different authority boundary.

Investigating a technician workstation does not replace reviewing server extension activity. Disabling an extension does not establish that unauthorized clients were removed. Finding a malicious client does not prove how the initial management access was obtained.

Security operations should correlate these paths rather than flattening them into one generic ScreenConnect alert.

Why a Current ScreenConnect Server May Still Be Exposed

One of the more significant lessons from the November campaign is the difference between core product version compliance and extension security.

Huntress observed unauthorized deployment activity involving current ScreenConnect server releases, including builds in the 26.6.6 and 26.7.6 families.

That does not establish a vulnerability in those server builds.

The suspected initial exposure involved third-party extensions. A current server can still host an extension whose authentication mechanism is unsafe.

Separately, a current hosted management instance can communicate with a technician workstation that has an older vulnerable host client.

These are different version and control boundaries.

A ScreenConnect readiness report should therefore record the following independently:

ComponentVersion or state to establishWhy it matters
Hosted or on-premises management instanceCurrent supported build and security configurationCore product lifecycle and administration
RMM+ and listed related extensionsInstalled, enabled, disabled, or removedIndependent extension API exposure
Technician host clientsActual installed version, at least 26.6.5CVE-2026-84869 protection
Endpoint access agentsInstalled version and approved management instanceSupportability and endpoint-side trust
Remote-access servicesExecutable provenance and relay destinationsUnauthorized secondary access
Active administrative sessionsIdentity and session validityPossible session hijacking or retained authority
Pending commands and Toolbox itemsApproved and unexpected future executionReinfection or unintended work after reconnection

The organization should also identify who owns each component.

A supplier may operate the hosted server, but the MSP remains responsible for its technician workstations and managed customers. An enterprise customer may own endpoint security while the MSP controls the remote-support application.

Those responsibilities must be explicit during containment.

AI-Generated Code Is Not Proof of an Autonomous AI Campaign

Huntress reported that portions of the RelayTrap C# source appeared to have been generated by a large language model, based on detailed explanatory comments and code structure.

That observation is consistent with possible AI assistance during development.

It is not evidence that SCKit performs model inference during execution, that ARTEX or CAIRN orchestrated this campaign, or that a model autonomously selected the victims.

The observed ScreenConnect activity can be explained through conventional remote-command execution, stolen or abused management authority, scripts, scheduled tasks, browser-session access, and unauthorized remote-support clients.

AI assistance during malware development and autonomous control of an intrusion are separate claims.

This article belongs in the wider AI cyberattack series because the distinction matters to enterprise threat assessment. The controls recommended here are justified by the observed attack paths, regardless of whether an LLM helped write the toolkit.

Common Response Failures

There are several ways to produce an apparently successful remediation while leaving meaningful risk behind.

Patching the Console but Not the Technician

The hosted console shows a current version, but a technician workstation continues using an older host client.

The CVE-2026-84869 execution path may remain available to that workstation.

The remedy is version verification on the endpoint itself, including clients that were previously installed or used through older sessions.

Disabling RMM+ and Assuming Every Endpoint Is Clean

The vulnerable API route is closed, but an attacker-controlled ScreenConnect client has already been installed.

That client may communicate with a different relay, independently of the original extension.

The remedy is endpoint investigation, removal of unauthorized remote-access software, and validation of persistence.

Resetting Passwords Without Revoking Sessions

Administrative passwords are changed, but active web or identity-provider sessions remain valid.

The SCKit toolkit was designed to target authenticated browser sessions, making session invalidation a material part of containment.

Use the applicable session-revocation procedures from a clean administrative environment and verify their effect.

Reconnecting for Troubleshooting

A technician opens an interactive ScreenConnect session to inspect a suspected infected machine.

That may trigger an unsafe interaction the toolkit was designed to exploit.

Use an independently trusted response path until the affected endpoint and technician-client risk have been addressed.

Removing Visible Malware Without Checking Re-creation

The suspicious service or task is removed, but another component or queued operation can recreate it.

Microsoft documented SCKit cleanup and redeployment behavior. Do not assume that a malware-provided cleanup script is a trustworthy eradication tool.

Verify the resulting state after controlled restart or reconnection, using appropriate monitoring and known-good configuration.

Reporting Requests as Compromises

A threat intelligence report identifies hundreds of thousands of file-execution requests.

The organization communicates that an equivalent number of technicians were compromised.

Those findings are not interchangeable.

The evidence must establish whether a vulnerable host connected at the relevant time and whether the action actually executed.

A Return-to-Service Gate for MSPs and Customers

Recovery should not begin merely because the vendor’s investigation has concluded or because the management console is available.

ConnectWise’s November 16 update marked its own investigation complete, while explicitly maintaining protective requirements for partner environments.

The supplier also warned that completion of its investigation did not establish that individual partner and customer systems were unaffected.

The MSP’s return-to-service decision therefore needs its own evidence.

The opening recovery diagram is a decision model, not an assertion that all stages have been completed in any particular organization.

For a return-to-service decision, require evidence of these outcomes:

  • The listed third-party extensions remain disabled, with the state verified across applicable management instances.
  • Technician clients meet the supported fixed-version requirement.
  • Unauthorized remote-access clients, tasks, services, and security exclusions have been investigated and remediated where found.
  • Potentially exposed administrative sessions and credentials have received the appropriate response.
  • Queued commands, shared tools, and automated workflows have been reviewed before reconnection.
  • Legitimate remote-support functionality works, and customer endpoint trust has an accepted disposition.

Retain an explicit record of devices or customers that could not be fully investigated.

The absence of telemetry is an investigation limitation, not evidence that a system was unaffected.

The Larger Architectural Lesson for Remote Management

ScreenConnect is a legitimate administration platform. Its value comes from enabling technicians to reach systems and perform operational work efficiently.

That same capability makes the management plane consequential during a compromise.

An MSP may operate one management environment that reaches many separate customer networks. A technician may have privileges across multiple customer tenants. A third-party extension may expose remote commands through an authentication mechanism separate from normal console access.

Those relationships should be treated as distinct trust boundaries.

A more resilient operating model should establish approved remote-management instances, inventory endpoint enrollment, constrain administrative privileges, govern extensions, maintain independently trusted response access, and protect the evidence needed to reconstruct management activity.

It should also address service continuity.

Disabling a remote-management extension or restricting technician connections can interrupt legitimate support. Business and technical owners need a fallback for urgent incidents, especially where managed customers depend on the MSP for security response.

A temporary support limitation may be preferable to exposing another privileged workstation, but that tradeoff should be a documented decision rather than an improvised reaction.

The lasting requirement is to ensure that a compromise in one remote-management component cannot automatically become unbounded administrative authority across the entire customer estate.

Conclusion

The November 2026 ScreenConnect incidents are a practical example of why security teams must investigate the actual control boundary rather than relying on a product name or version alone.

CVE-2026-127063 affects a third-party RMM+ extension. CVE-2026-84869 affects ScreenConnect clients. SCKit introduced unauthorized remote access and attempted additional propagation through technician sessions.

Those are related but separate problems.

Huntress documented substantial unauthorized endpoint deployment and separately confirmed technician-client exploitation. It did not confirm that the reconstructed SCKit endpoint-to-technician-to-console propagation chain succeeded. ConnectWise completed its own investigation on November 16 but continued to require extension disablement, client updates, and partner-specific investigation.

The operational response has to follow those boundaries.

Disable the specified extensions. Verify technician host-client versions. Preserve and investigate management activity. Remove unauthorized endpoint access. Revoke exposed sessions. Validate customer service before restoring normal remote support.

A current ScreenConnect console is not proof that the management estate is secure. The meaningful result is evidence that unauthorized command paths are blocked, technician workstations are protected, and previously established attacker access has been removed.

For the broader eradication and evidence model, use the persistent-access recovery guide. The ConnectWise Cloud Backup investigation covers the separate application-trust incident and its customer-tenant response.

External References

The post ScreenConnect, SCKit, and CVE-2026-84869: What to Check appeared first on Digital Thought Disruption.