Malicious npm package @guicai777/mcp-doctor posed as an MCP diagnostic tool and dropped a hidden Windows executable, with a README telling users to run mshta (MAL-2026-20295)
Version 1.0.1 of the npm package @guicai777/mcp-doctor was presented as an MCP diagnostic utility. According to the Amazon Inspector analysis in OSV, it ships an ~866 KB bridge.bin that is a base64-encoded, XOR-obfuscated Windows executable. On Windows, index.js decodes it to %LOCALAPPDATA%\mcp-doctor\mcp-bridge.exe and starts it detached and hidden, after first checking https://fix-mcp-bridge.com/v.json as a remote kill switch. The README also tells users to run mshta https://fix-mcp-bridge.com/init, which downloads and runs remote script from the same domain. npm has since replaced the package with a security placeholder.
What it is
Reported to OSV on 20 November 2026 by Amazon Inspector (IN-MAL-2026-022951) and GitHub's malware feed (GHSA-h82f-cjm4-g22r). Affected version: 1.0.1.
The payload is unsigned and unverified; the encoding only serves to hide it from scanners and reviewers.
GitHub's guidance: any computer that installed or ran the package should be considered fully compromised, and secrets on it rotated from a different machine.
How to check
- Search lockfiles and global installs for @guicai777/mcp-doctor.
- On Windows, look for %LOCALAPPDATA%\mcp-doctor\mcp-bridge.exe and for network connections or DNS lookups to fix-mcp-bridge.com.
- Check shell history for mshta commands pointing at fix-mcp-bridge.com.
How to fix
- Remove the package, and treat any machine that ran it (or the mshta command) as compromised: reimage it if you can.
- Rotate every secret that machine held (API keys, SSH keys, browser sessions, cloud credentials) from a clean device.
- Block fix-mcp-bridge.com, and only install MCP tooling from publishers you can verify.
grep -rn '@guicai777/mcp-doctor' --include=package-lock.json --include=pnpm-lock.yaml --include=yarn.lock .
npm ls -g @guicai777/mcp-doctorSources
- criticalMalicious npm packages posing as AI SDKs and MCP servers: character-ai-sdk, char-ai-sdk, @mibbp/agent-observer, mrnima-openai-chatgpt-bot and @fdcasj_dfew/md-doc-mcp
- criticalMalicious AI agent and MCP packages: mcp-aevo-server (PyPI) steals Aevo trading keys and wallet private keys; vamp-mcp, tsc-ai and @tnhorg/agent-skills (npm) flagged as malware
- criticalCVE-2026-77359: Jupyter MCP Server's /api/connect and /api/stop routes skipped the token and Origin checks, so a website or network caller could repoint it to an attacker's Jupyter backend (fixed 1.0.3)
- criticalCVE-2026-76807: Doist's official Todoist MCP HTTP server let any website use DNS rebinding to read, create and delete your Todoist data (fixed 10.3.3)
- criticalMalicious npm packages posing as MCP servers and AI APIs: @shadowdu/memos-mcp, @shadowdqj/image-mcp, mem0-mcp-server and ai-apis
- criticalMalicious npm and PyPI packages posing as AI and MCP tools: depfix-ai, agent-trace-kit, @go-gost/mcp-darwin-x64 and my-mcp-server-driving-license-ocr