AI Nexus HubAI Nexus Hub
PostsQuestionsDiscussions

Command Palette

Search for a command to run...

AI Nexus Hub

CONTACT:
axc@cyberlogitec.com

Content
PostsQuestionsDiscussionsTopics
Stay in the loop

Weekly digest of the best AI posts. No spam.

© CLT 2026. All rights reserved.
Privacy
Posts  /  Claude-code  /  Use the Claude Code extension in Cursor? Check your machine for the 2.1.312 / 2.1.313 backdoor in 2 minutes
0
Claude-code

Use the Claude Code extension in Cursor? Check your machine for the 2.1.312 / 2.1.313 backdoor in 2 minutes

HL
Hung Lay
Nov 2, 2026 · 3 min read
Banner: Claude Code extension in Cursor? Check for the 2.1.312 / 2.1.313 backdoor in three checks: version, SSH key, persistence

TL;DR

  • On 2026-10-29, malicious builds of the Claude Code extension (versions 2.1.312 and 2.1.313) were on the Cursor marketplace for about 11 hours.

  • They tried to plant an SSH key and a reverse tunnel. Updating or uninstalling does not remove that.

  • Three copy-paste checks tell you if your machine is clean: about 2 minutes.

What it is

If you use Cursor, it installs VS Code extensions from its own marketplace, not from Microsoft's. According to Cursor's security notice, someone published modified builds of Anthropic.claude-code 2.1.312 and 2.1.313 there, for macOS and Windows. Cursor says the builds "were not developed by Anthropic", and it estimates up to 5,000 users may be affected (as reported by YourStory).

The bad code ran as soon as Cursor loaded the extension. No click needed. According to Cursor and the Phorion advisory, it did four things:

  1. Sent a beacon: hostname, user name, OS, public IP, git email, and the names (not values) of environment variables like *TOKEN* or *SECRET*.

  2. Added an attacker SSH key labeled svc-telemetry to authorized_keys.

  3. Opened a reverse SSH tunnel to 134.209.82.225, retrying every 60 seconds.

  4. Made itself persistent: scheduled task SysMon on Windows, LaunchAgent com.apple.sysmond on macOS.

Cursor says the SSH parts usually fail without admin rights, and the attacker server is now offline. The beacon still went out.

Illustration: what the malicious build did (beacon, SSH key, reverse tunnel, persistence) next to the three checks you run

Illustration

Try it in 5 minutes

Prerequisites: a terminal on the machine where you use Cursor. No install needed.

Step 1: See which Claude Code versions Cursor installed.

# Windows PowerShell
Get-ChildItem $HOME\.cursor\extensions -Filter anthropic.claude-code*
# macOS / Linux
ls ~/.cursor/extensions | grep anthropic.claude-code

A folder named anthropic.claude-code-2.1.312-... or ...-2.1.313-... means you had a bad build. Run step 2 anyway: an auto-update may have removed the old folder.

Step 2: Search for the attacker's SSH key (the check from Cursor's notice).

# Windows PowerShell
Select-String -Path "$HOME\.ssh\authorized_keys" -Pattern svc-telemetry
Select-String -Path "C:\ProgramData\ssh\administrators_authorized_keys" -Pattern svc-telemetry
# macOS / Linux
grep -qs svc-telemetry ~/.ssh/authorized_keys && echo "KEY FOUND" || echo "not impacted"

Step 3: Read the result, then check for persistence.

# Windows PowerShell
Get-ScheduledTask -TaskName SysMon -ErrorAction SilentlyContinue
Test-Path "$env:APPDATA\Microsoft\Windows\Start Menu\Programs\Startup\sysmon.cmd", "$HOME\.config\smt.cmd"
# macOS
ls ~/Library/LaunchAgents/com.apple.sysmond.plist ~/.config/.smt 2>/dev/null

What you will see: no output, "not impacted" and False mean you are not affected. Found anything? Disconnect, tell your security team, run Cursor's clean-up script (linked in its notice), and rotate the keys and tokens on that machine.

Tip: Update the extension to 2.1.314 or later. According to Cursor and the published reports, the builds on the Microsoft VS Code Marketplace, Open VSX and the npm CLI were not affected. Open VSX shows 2.1.316 today.

Caveats

  • Updating is not cleaning. Cursor warns that uninstalling "may not be sufficient". The key and the scheduled task stay until you remove them.

  • Root cause is not public. Neither Cursor nor Anthropic has explained how the builds reached the marketplace.

  • Read scripts before you run them. Cursor's Windows script needs an Administrator PowerShell. Open it and read it first.

  • Env var names leaked even on clean machines. If you had a bad build, assume the attacker knows which secrets you have, and rotate them.

  • VS Code users: you are not in scope, according to Cursor. The checks still take 2 minutes.

Claude Quick Hits

  • Claude Code 2.1.316 (2026-10-31): CLAUDE_CODE_SUBAGENT_EFFORT sets the effort level for subagents; CLAUDE_CODE_DISABLE_ULTRATHINK turns the keyword off.

  • Fast mode for Opus 4.8 is deprecated (2026-10-30): removal on 2026-11-30. Move fast-mode calls to Opus 5.5.

AI Tool Radar

Repo/Tool

What it does (1 line)

Stars (growth)

Quick try

microsoft/TokenEconomizer

Plugin that learns from past sessions to cut coding-agent cost (18.4–30.9% on SWE-Bench, according to the README)

3,968 (new, created 2026-10-13)

claude plugin install tokeneconomizer@tokeneconomizer after adding the clone as a marketplace

haydenschmitty/floorplan

Claude Code or Codex draws an interactive map of your codebase

1,818 (+1,425 since 2026-10-30)

claude plugin marketplace add haydenschmitty/floorplan then claude plugin install floorplan@floorplan

Sources

  • https://cursor.com/blog/claude-code-extension-security-notice

  • https://phorion.io/blog/offical-claude-code-vsx-compromise/

  • https://news.ycombinator.com/item?id=50300683

  • https://yourstory.com/ai-story/cursor-warns-of-malicious-extension-affecting-up-to-5000-users

  • https://open-vsx.org/extension/Anthropic/claude-code

  • https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md

  • https://platform.claude.com/docs/en/release-notes/overview

  • https://github.com/microsoft/TokenEconomizer

  • https://github.com/haydenschmitty/floorplan

#claude-code#ai#research
Y
Be kind. Markdown supported. · @ to mention
What it isTry it in 5 minutesCaveatsClaude Quick HitsAI Tool RadarSources

Related posts

Lock down any MCP server or agent command with one word: Anthropic's srt sandbox, no container needed

TL;DR srt is Anthropic's open-source sandbox CLI, the same engine that Claude Code's /sandbox mode is built on. Put srt in front of any command or local MCP ser

HLHung LayOct 29, 20261 min
0 0 1

Claude Code on Bedrock or Vertex with a custom model name? Keep the full system prompt after the 2.1.312 update with one setting

TL;DR What it is: Claude Code 2.1.312 (released 2026-10-27) now sends a shorter system prompt when your model name is one it does not recognize on Bedrock, Vert

HLHung LayOct 29, 20261 min
0 0 0

Turn Claude Code into a security auditor that checks its own findings: Cloudflare's security-audit skill, installed in 2 minutes

TL;DR What it is: cloudflare/security-audit-skill is one agent skill that makes Claude Code run a six-phase security audit, where a different agent tries to dis

HLHung LayOct 28, 20261 min
0 0 4

Stop copying setup snippets: browse 315 Claude Code plugins in the official directory and install the right one in 2 minutes

TL;DR What it is: claude-plugins-official is the catalog behind Claude Code's /plugin browser. It lists 315 plugins: 39 by Anthropic, 276 by partners and other

HLHung LayOct 28, 20261 min
0 0 1

More from Hung Lay

Lock down any MCP server or agent command with one word: Anthropic's srt sandbox, no container needed

TL;DR srt is Anthropic's open-source sandbox CLI, the same engine that Claude Code's /sandbox mode is built on. Put srt in front of any command or local MCP ser

HLHung LayOct 29, 20261 min
0 0 1

Claude Code on Bedrock or Vertex with a custom model name? Keep the full system prompt after the 2.1.312 update with one setting

TL;DR What it is: Claude Code 2.1.312 (released 2026-10-27) now sends a shorter system prompt when your model name is one it does not recognize on Bedrock, Vert

HLHung LayOct 29, 20261 min
0 0 0

Turn Claude Code into a security auditor that checks its own findings: Cloudflare's security-audit skill, installed in 2 minutes

TL;DR What it is: cloudflare/security-audit-skill is one agent skill that makes Claude Code run a six-phase security audit, where a different agent tries to dis

HLHung LayOct 28, 20261 min
0 0 4

Stop copying setup snippets: browse 315 Claude Code plugins in the official directory and install the right one in 2 minutes

TL;DR What it is: claude-plugins-official is the catalog behind Claude Code's /plugin browser. It lists 315 plugins: 39 by Anthropic, 276 by partners and other

HLHung LayOct 28, 20261 min
0 0 1