Skip to content
RESEARCH INDEX BREACHROAD / INTELLIGENCE NOTE

ScreenConnect and SCKit: RMM+ Extension Attacks

ConnectWise closed its ScreenConnect investigation while attacks continue. We explain SCKit, CVE-2026-84869, RMM+ extensions and safe MSP response.

PUBLIC RESEARCH
AUTHOR
/ CEO Breachroad · OSCP · PNPT
PUBLISHED
17 November 2026
READING TIME
22 min read
TOPIC
Supply Chain Security
ScreenConnect and SCKit: RMM+ Extension Attacks

ConnectWise has completed its investigation into activity targeting ScreenConnect environments, but protective work remains necessary. Researchers continue to observe activity against on-premises installations where identified third-party extensions have not been disabled or technician clients remain outdated.

The final update was posted at 20:00 UTC on 16 November 2026. Administrators, MSPs and security teams should therefore treat the case on 17 November as an active operational issue rather than closed history. The official ConnectWise Trust Center advisory requires two distinct actions: disable the named RMM+ extensions and verify that every technician client runs version 26.6.5 or later.

Updating only the server does not prove that technician workstations are protected. Disabling an extension does not remove access or persistence an attacker may already have established. Those are the two most important traps in this incident.

What ConnectWise confirms

ConnectWise confirms reports of unauthorised access to some partner ScreenConnect accounts and use of remote-access functions to deploy unauthorised software. Its investigation found no evidence that ConnectWise infrastructure was compromised or that initial access resulted from a vulnerability in the ScreenConnect server itself.

The company’s assessment points to abuse of independently developed extensions in the RMM+ family. Approximately 400 partner environments had at least one relevant extension installed. That figure represents extension presence, not 400 confirmed compromises.

The extensions that must remain disabled are:

  • Labtech Plus;
  • RMM Plus;
  • RMM+;
  • RMM Plus Password Link;
  • RMM+ Password Link.

ConnectWise disabled them as a precaution in monitored cloud environments, but it cannot perform the same action for on-premises administrators. The extensions must not be re-enabled or reinstalled pending further guidance.

The extension CVE and a separate client flaw

At least two distinct issues appear in this case and should not be collapsed into one CVE.

The first involves the independent RMM+ extension. ConnectWise reports an authentication flaw in versions 1.0.31 and 1.0.32 that could accept an empty access key when no key had been configured. The issue is assigned CVE-2026-127063, with disclosure material linked from the Trust Center.

The second is CVE-2026-84869 in the ScreenConnect client. The official ConnectWise security bulletin describes missing authorisation that could allow files to be transferred and executed through an active session without authorisation or host confirmation under certain conditions. Client releases before 26.6.5 are affected, and the CVSS score is 9.9.

Version 26.6.5 blocks the observed SCKit attempt to move from an infected endpoint to a technician workstation. It does not fix the extension’s initial-access issue and cannot remove a compromise that already occurred. Conversely, disabling an extension does not update clients.

What SCKit means in this campaign

According to findings summarised by ConnectWise, independent researchers observed a toolkit called SCKit. After gaining access, actors deployed unauthorised ScreenConnect clients and additional scripts. The toolkit was designed to attempt propagation from a compromised endpoint to a technician workstation during an interactive connection and then onward to other systems.

ConnectWise emphasises that researchers had not confirmed successful execution of the endpoint-to-technician technique at publication time. They did confirm that client 26.6.5 and later blocks it. Accurate analysis must retain both facts: the toolkit’s designed capability and the absence of confirmed success for that specific technique in observed telemetry.

The risk to an MSP environment remains high. A legitimate remote-management platform operates at scale. One account, extension or technician workstation may connect to many customers, producing a wider blast radius than an ordinary single-enterprise application.

Why a current server is insufficient

ScreenConnect includes the server, technician clients and access agents on endpoints. An operator may see an up-to-date server or use a cloud service automatically upgraded by the vendor while an older client remains on a technician laptop.

ConnectWise explicitly requires verification of the installed client version on every technician workstation. The console version is not enough. Host clients may need reinstallation and access agents may need updating after the server change.

Inventory should include:

  • every cloud and on-premises instance;
  • every technician workstation, including standby devices;
  • clients preserved in VDI images;
  • access agents on customer endpoints;
  • Automate integration and other RMM tools;
  • inactive and emergency accounts;
  • extensions on each instance;
  • versions contained in backups and machine templates.

This is a supply-chain problem: the version of the central product does not describe the whole ecosystem performing trusted operations.

Immediate action for MSPs

The first phase does not require proof of compromise. An on-premises administrator should confirm the state and disable the five named extensions immediately. Next, check every technician client and upgrade to 26.6.5 or a later supported version.

Then:

  1. preserve Run Command history, extension API activity, Toolbox and audit logs;
  2. inventory unauthorised remote-access clients and newly deployed tools;
  3. review administrators, roles, permissions and configuration changes;
  4. revoke active sessions from a trusted device;
  5. reset relevant credentials and enforce MFA;
  6. remove unauthorised queued commands and tools only after preserving evidence;
  7. isolate suspicious devices;
  8. contact managed customers through the established incident plan.

A password reset does not always invalidate an existing session. Explicitly revoke tokens and sessions in ScreenConnect and the connected identity provider.

Do not investigate through suspected ScreenConnect access

ConnectWise advises against using ScreenConnect to investigate or clean an endpoint suspected of compromise. When the remote-management channel is part of the attack path, actions through the same channel may be observed, modified or used for further movement.

Investigators should use a trusted out-of-band capability: EDR response, physical access, a separate jump host or another independently verified channel. Preserve images, memory, logs and chronology before removing persistence.

This principle is broader than one product. An administration platform involved in an incident becomes evidence and a potentially untrusted medium. Security monitoring for MSPs should include an independent response path.

Telemetry to review

Priority sources are ScreenConnect audit logs, command history, extension activity, client-installation events, Windows processes and services, EDR, DNS, proxy and identity-provider logs. In customer environments, look for unknown remote-access tools and processes launched by an agent.

Correlate signals such as:

  • a command downloading a script or installer from an unknown domain;
  • silent installation of another remote-access client;
  • a new administrator or role change without a ticket;
  • technician login from an unusual location or device;
  • an empty or unexpected RMM+ API key value;
  • the same command run across many customers;
  • a connection to a technician workstation immediately after work on a suspicious endpoint;
  • EDR disablement, policy changes or log removal;
  • a new scheduled task, service or persistence under a user profile;
  • a session that remains active after a password change.

The absence of a known IOC does not exclude compromise. Legitimate RMM features let an actor operate without custom malware. Behaviour, identity and state change therefore matter more than a file signature alone.

Blast-radius modelling for service providers

An MSP should build a matrix of technician account → instance → customers → endpoints → permissions. Mark which resources were reachable during the suspicious period. Not every account can access every customer, and an accurate map narrows investigation and notification without omitting dependencies.

Indirect connections matter too. A technician may have credentials in the browser, access to documentation, a password vault, vendor portals and email. Workstation compromise does not end with ScreenConnect. Third-party risk management should model these dependencies.

Customers deserve concrete answers: was their endpoint in scope, which logs were examined, what actions were performed and which uncertainties remain. “We updated the server” does not answer those questions.

Hardening after containment

After investigation, reduce the extension count, remove plugins without owners and maintain an allowlist. Every extension needs a source, version, owner, permission scope, review result and retirement plan.

Technician accounts require phishing-resistant MFA, separate administrative identities and just-in-time access. Do not use the same account for email, everyday browsing and customer administration. Technician devices should be privileged workstations with application control and restricted Internet access.

Separate deployment from observation. If an actor controls the RMM, they should not be able to erase independent SIEM logs or disable every sensor. Export audit data outside the tenant and alert on retention changes.

Updating is not incident eradication

A patch removes a known path for future abuse. It does not remove an account, agent, scheduled task, token, key or remote tool already created. ConnectWise explicitly warns that disabling extensions and updating clients do not remove unauthorised access established on endpoints.

Use a controlled sequence:

  1. preserve evidence;
  2. isolate confirmed or strongly suspected systems;
  3. close initial access paths;
  4. revoke sessions and credentials;
  5. remove persistence or rebuild from trusted media;
  6. update server, clients and agents;
  7. retest and hunt across the wider scope;
  8. restore connectivity in stages.

For a high-trust technician workstation, a rebuild may provide greater assurance than selective cleaning.

A plan for customers of an MSP

Customers do not need to wait passively. Ask whether the provider uses ScreenConnect, had the named extensions, verified technician-client versions and reviewed telemetry for unauthorised commands.

The customer’s SOC should search for new remote-access software, unexpected processes launched by an agent, local and domain account changes and unusual traffic from management workstations. If the provider used privileged credentials in the customer environment, evaluate rotation after containment.

Contracts and playbooks should define notification timing, log access, evidence preservation and an emergency channel independent of RMM. Our guide to software supply-chain attacks explains why those clauses are technical controls rather than paperwork.

Source facts and Breachroad conclusions

ConnectWise confirms completion of its investigation, continuing activity against unremediated environments, the need to disable extensions, the requirement to update every technician client to at least 26.6.5 and the importance of independent investigation. It also confirms roughly 400 environments with extensions—not 400 victims—and no evidence of compromise in ConnectWise infrastructure.

The blast-radius matrix, response sequence, additional correlations and architecture recommendations are Breachroad conclusions. Organisations can validate their controls through infrastructure penetration testing and an IT security audit.

The key takeaway

ScreenConnect operators need to complete two separate jobs: disable the named extensions and update the client on every technician workstation. They must then assume that a patch cannot reverse time and investigate whether access or persistence existed beforehand.

The largest risk comes from transitive trust: customer endpoints trust technicians, technicians trust the platform, and the platform trusts an extension. Without an inventory of the complete chain, a current server version can create false confidence.

SHARE / COPY