Cursor Claude Code extension backdoor Oct 29 2026 — Current Guide
Cursor Claude Code extension backdoor Oct 29 2026: Cursor Claude Code extension backdoor Oct 29 2026: up to ~5,000 devs exposed; step-by-step detection, remova…

View article sections
- 01Cursor Claude Code extension backdoor Oct 29 2026
- 02Confirmed timeline and scope
- 03What the malware did (technical summary)
- 04Who should be worried right now
- 05Immediate steps to contain and remediate (step-by-step)
- 06Risk assessment and likely impacts
- 07What to change in your security posture
- 08Quick comparison: actions by priority and timeframe
- 09Confirmed and independent sources
- 10Practical next steps for small dev teams and contractors
- 11FAQs
- 12Closing summary and final advice
- 13Related guides and resources
- 14Frequently asked questions
- 15Need practical help?
- 16Topic in context
- 17Sources and further reading
Cursor Claude Code extension backdoor Oct 29 2026
This dated, source-based update separates verified details, limitations, and practical next steps.
What changed (confirmed): On Oct 29, 2026 Cursor disclosed that a maliciously modified build of Anthropic.claude-code (versions 2.1.312 and 2.1.313) was distributed through the Cursor VS Code Extension Marketplace for roughly 11 hours and may have affected up to ~5,000 users. This incident matters because the tampered package installed a persistent SSH backdoor on developer machines, allowing remote access to potentially sensitive code and systems [1].
Summary: This article explains the verified facts, technical impact, and step-by-step remediation and audit guidance for developers and small ops teams. Action is time-sensitive for any team that installed or used the extension between the reported window. Where statements are confirmed, independent, or investigative, they are labeled.
Confirmed timeline and scope
Confirmed (official): Cursor published a security notice on Oct 29, 2026 describing that Anthropic.claude-code versions 2.1.312 and 2.1.313 were maliciously modified and distributed from its VS Code Extension Marketplace for about 11 hours [1].
Independent reporting: Security researchers and incident responders who examined affected machines reported the build delivered a persistent SSH backdoor and called the package trojanized. These analyses were published by StepSecurity and pwn.watch on Oct 30, 2026 and corroborate the behavior described by Cursor [2][3].
Estimated impact (confirmed by Cursor): Cursor estimated the tampered package may have reached up to ~5,000 users during the distribution window [1]. As of Nov 1, 2026, that count and scope remain the best public estimate; the exact number of compromised hosts is subject to ongoing investigation [1].
What the malware did (technical summary)
Independent analyses found the malicious builds altered the extension’s installation or post-install scripts to place an SSH key and a small daemon or system service on host machines. That artifact created an open, persistent access method over SSH so a remote operator could connect to affected machines without the owner’s credentials [2][3].
Confirmed elements:
- Malicious builds: Anthropic.claude-code 2.1.312 and 2.1.313 distributed via Cursor’s marketplace [1].
- Persistent access method: an SSH backdoor implanted on developer workstations, enabling remote shell access [2][3].
- Distribution window: approximately 11 hours on Oct 29, 2026 [1].
How attackers typically persist an SSH backdoor
While every trojan differs, common persistence techniques seen in this incident and similar supply-chain compromises include creating or appending to authorized_keys, installing a private key in a hidden directory, adding a system-level service or cron job to relaunch a small listener, and disguising files as legitimate extension assets. The independent reports noted patterns consistent with those techniques in affected builds [2][3].
Who should be worried right now
If you installed or updated the Anthropic.claude-code extension from Cursor’s VS Code Extension Marketplace during the affected window on Oct 29, 2026, consider your machine and any networks it connects to at elevated risk [1].
Priority groups:
- Developers and dev teams who run code or have SSH access to servers from their development workstations.
- CI/CD hosts, build agents, or shared dev machines that may have been used while the malicious extension was installed.
- Small businesses and contractors who store secrets, SSH keys, or credentials locally.
Immediate steps to contain and remediate (step-by-step)
Below are prioritized actions. These are practical, conservative steps for developers and IT teams. Labeling: the removal steps below are recommended mitigation actions based on independent reporting and common incident response practice; they do not replace an organization’s formal IR process.
- Uninstall the extension immediately. Remove Anthropic.claude-code from VS Code and from any machines where it was installed during the affected window [1].
- Isolate affected hosts. Disconnect suspected machines from networks to prevent outbound/inbound access, or place them on a segregated VLAN for investigation.
- Detect and remove implanted SSH keys. Check user and system SSH authorized keys and known key locations (~/.ssh/authorized_keys, /root/.ssh/authorized_keys, hidden folders). Remove unfamiliar keys and any files with unusual permissions. Independent analysis found implanted keys and a persistent agent in affected systems [2][3].
- Rotate credentials and keys. Immediately rotate SSH keys used from affected hosts, as well as API tokens and any credentials stored locally. Assume any credential accessible from an infected host may be compromised.
- Collect forensic data. Save memory and disk snapshots, copy logs, and document installed files and services before making destructive changes. This helps incident response and attribution.
- Check for new services, cron jobs, or scheduled tasks. Look for suspicious systemd units, init scripts, launchd plists (macOS), or cron entries that reintroduce a backdoor.
- Scan and rebuild if needed. For any host that showed evidence of backdoor activity, consider rebuilding from known-good images or reinstalling the OS after preserving forensic artifacts. Rebuilding removes hidden persistence with the highest assurance.
- Audit CI/CD pipelines and servers. Ensure build agents, deployment keys, and server-side access were not used or replaced by the attacker. Rotate keys and secrets used by CI systems.
- Notify stakeholders. Inform internal security teams, clients, and any parties that rely on compromised machines. Follow contractual and regulatory incident disclosure requirements where applicable.
- Report findings to Cursor and Anthropic. Cursor has posted an official security notice; report new indicators you find to them and to Anthropic to assist containment [1].
Quick detection checklist (command-style hints)
- Look for unexpected SSH keys: inspect ~/.ssh and /root/.ssh for new files and authorized_keys entries.
- List recent package installs and extension directories: check VS Code’s extensions folder for modified timestamps during Oct 29, 2026.
- Search for persistent processes and system services added on or after Oct 29, 2026.
- Review outbound connections and unknown listening ports created by user processes.
Risk assessment and likely impacts
Confirmed: The injected SSH backdoor creates a high-impact confidentiality and integrity risk for affected machines because remote operators could access code, secrets, and networked systems from the developer endpoint [2][3].
Estimated typical consequences (analysis and estimates): Remote access from a compromised developer workstation can lead to credential theft, source-code exfiltration, unauthorized deployments, and lateral movement to production or staging servers. The actual damage depends on whether the attacker exploited access to reach other high-value targets. As of Nov 1, 2026, public reporting does not document large-scale downstream breaches tied to this supply-chain incident [1][2][3].
What to change in your security posture
Supply-chain compromises like this highlight several controls that materially reduce risk over time:
- Use least-privilege SSH keys and avoid reusing keys across machines and services.
- Keep secrets out of local developer workstations; prefer ephemeral credentials, secret managers, and short-lived tokens.
- Harden developer endpoints: EDR (endpoint detection and response), managed updates, and host-based auditing can speed detection.
- Use code signing and verified extension repositories where possible. Verify checksums and release provenance for critical tools.
Quick comparison: actions by priority and timeframe
| Priority | Action | Timeframe | Why |
|---|---|---|---|
| Critical | Uninstall extension; isolate hosts | Immediate | Stops further access and limits spread |
| High | Rotate SSH keys and credentials | Within 24 hours | Prevents reuse of compromised keys |
| High | Collect forensic data and scan | 24–72 hours | Determines scope and supports recovery |
| Recommended | Rebuild compromised hosts | As soon as feasible | Most reliable way to remove hidden persistence |
Confirmed and independent sources
Official disclosure: Cursor’s security notice confirming the malicious builds and the distribution window is the primary, confirmed source for the incident timeline and scope [1].
Independent analysis: StepSecurity and pwn.watch published technical write-ups and detection findings that corroborate the presence of a persistent SSH backdoor in the tampered extension builds [2][3].
Note: As of Nov 1, 2026 these sources represent the verified public record. Ongoing investigations by Cursor and other responders may add new indicators.
Practical next steps for small dev teams and contractors
- Determine which team members installed or updated anthopic.claude-code on Oct 29, 2026.
- Follow the containment checklist above for each suspected host.
- Prioritize servers and CI systems with credentials that were accessible from compromised hosts.
- Engage an incident response partner if you lack forensic expertise or if the incident scope is high.
- Document all actions and communications for compliance and insurance purposes.
FAQs
Q: I installed Anthropic.claude-code on Oct 29 — how do I know if my machine was compromised?
A: Check for unknown SSH keys or entries in ~/.ssh/authorized_keys and for new system services or background processes. If you see unknown authorized keys or an unfamiliar daemon that started after the install, treat the host as compromised and follow the containment steps above. Independent reports show implanted keys and persistent agents in affected systems [2][3].
Q: Do I need to rotate all SSH keys or only the ones used from the infected machine?
A: Rotate any SSH keys and credentials that were present or used on the infected host. If you reused the same key across multiple machines or services, rotate those too. Assume any credential accessible from the infected host may be exposed.
Q: Can I remove the malicious files and keep using the same machine?
A: Removing files may eliminate visible artifacts, but sophisticated persistence can survive simple cleanup. For hosts with confirmed compromise, the safest option is a full rebuild from trusted media after preserving forensic evidence if needed.
Q: Has Cursor fixed the marketplace and is it safe now?
A: Cursor posted an official notice about the tampered builds and alerted users [1]. As of Nov 1, 2026, Cursor is the authoritative contact for the marketplace remediation status. Follow Cursor’s guidance and verify that future extension versions are signed and vetted before reinstalling.
Q: Should I stop using VS Code extensions from third-party marketplaces?
A: Third-party extensions increase supply-chain risk. Use extensions from trusted publishers, enable extension signing if your environment supports it, and limit extension use on high-privilege machines. For critical workflows, adopt hardened developer images and use ephemeral dev environments.
Closing summary and final advice
Confirmed details show that a tampered Anthropic.claude-code build distributed via the Cursor VS Code Extension Marketplace on Oct 29, 2026 installed an SSH backdoor on some developer machines and may have affected up to ~5,000 users [1][2][3]. If you installed or updated the extension during the reported window, treat hosts as potentially compromised, isolate them, and follow the containment, detection, and recovery steps above.
For small teams, the practical approach is immediate containment, rotating credentials, and rebuilding any host with confirmed indicators of compromise. For larger organizations, coordinate with incident response and legal teams while documenting actions and communications. Finally, learn from the incident: reduce local secrets exposure, enforce least privilege, and strengthen extension and package verification into your developer workflow.
Frequently asked questions
I installed Anthropic.claude-code on Oct 29 — how do I know if my machine was compromised?
Check for unknown SSH keys in ~/.ssh/authorized_keys and for new system services, cron jobs, or background processes that started after the install. If you see unfamiliar keys or daemons, isolate the host and follow the containment steps: uninstall the extension, preserve logs and artifacts, and consult incident response. Independent analyses reported implanted keys and persistent agents in affected builds [2][3].
Do I need to rotate all SSH keys or only those used from the infected machine?
Rotate any SSH keys and credentials that were present on the infected host and any credentials reused across systems. Assume any secret or key accessible from a compromised developer machine may be exposed and rotate accordingly.
Can I safely remove the malicious files and continue using the same machine?
Simple file removal can remove visible artifacts but may not eliminate all persistence. For machines with confirmed compromise, the most reliable remediation is rebuilding the OS from trusted media after preserving forensic data if needed. At minimum, isolate the host, rotate credentials, and perform a thorough audit before returning it to production use.
Has Cursor resolved the marketplace issue and is it safe to reinstall the extension?
Cursor published an official security notice confirming the tampered builds and the distribution window [1]. As of Nov 1, 2026, follow Cursor’s guidance for marketplace remediation and verify future extension versions are provided by trusted publishers and signed or checksummed before reinstalling.
What longer-term changes should teams make to prevent similar incidents?
Adopt least-privilege SSH practices, store secrets in centralized secret managers rather than local files, harden developer endpoints with EDR and strict update controls, and verify extension or package provenance. Consider ephemeral development environments to limit long-lived credentials on local machines.
Need practical help?
Fixit Solutions Inc. — Contact Fixit Solutions today to request a free estimate, schedule a repair or discuss your business technology needs. Service area: Lake Forest, CA.
Topic in context

Sources and further reading
These links were validated and checked when possible when this article was created; some publishers limit automated requests. Facts, guidance, prices, regulations, and availability can change.
- Security notice — malicious version of Anthropic.claude-code (2.1.312 and 2.1.313) identified in the Cursor VS Code Extension Marketplace — Cursor (official blog) (2026-10-29) — primary source
- Trojanized Claude Code Extension Delivered Persistent SSH Backdoor to Developer Machines — StepSecurity (2026-10-30)
- Trojanized Claude Code VS Code extension (Anthropic.claude-code 2.1.312 and 2.1.313) in Cursor's extension marketplace planted an SSH backdoor — pwn.watch (2026-10-30)

