Malicious code in @ikbal_fadilah_vanexa01/vanexa-agent (npm)
The npm package @ikbal_fadilah_vanexa01/vanexa-agent contained malicious code that establishes persistent remote command-and-control via a hardcoded Cloudflare Workers relay and exfiltrates user data to an author-controlled endpoint. The bundled daemon spawns arbitrary shell commands and silently relays chat messages and device identifiers when no API key is configured.
- Disclosed
- Last updated
- Blast radius
- Any host running the @ikbal_fadilah_vanexa01/vanexa-agent package; remote command execution and data exfiltration possible.
- Ecosystems
- Attack vectors
- Affected entities
- @ikbal_fadilah_vanexa01/vanexa-agentnpm package containing malicious code in bundled daemon
The npm package @ikbal_fadilah_vanexa01/vanexa-agent contained malicious code embedded in its bundled daemon (dist/bundle.cjs). The daemon opens an outbound WebSocket connection to a hardcoded relay at vanexa-agent-relay.hanazaki542.workers.dev, an author-controlled Cloudflare Workers subdomain.\n\nThe relay parses inbound JSON messages and dispatches 'task_request' messages to an agent exposing a 'terminal.exec' tool, which spawns arbitrary shell commands on the installer's host. The relay URL is not effectively user-configurable; attempts to redirect to alternative vanexa-agent-relay.workers.dev hosts are rewritten back to hanazaki542.workers.dev. This provides persistent remote command-and-control capabilities to anyone controlling the relay.\n\nAdditionally, when no API key is configured, the client posts full chat messages along with a stable device hash to a second hardcoded author-controlled endpoint (vanexa-ai-proxy.hanazaki542.workers.dev) instead of the user's chosen LLM provider. This silently relays user prompt content and device identifiers through the author's infrastructure without user consent.\n\nThe malicious package was identified by the OpenSSF malicious-packages project and assigned identifier MAL-2026-13364.
Indicators of compromise
- Packages
- @ikbal_fadilah_vanexa01/vanexa-agent
- Domains
- vanexa-agent-relay.hanazaki542.workers.dev
- vanexa-ai-proxy.hanazaki542.workers.dev
Remediation
- Immediately uninstall @ikbal_fadilah_vanexa01/vanexa-agent from all systems
- Audit any systems that had the package installed for unauthorized shell command execution or data exfiltration
- Review network logs for connections to vanexa-agent-relay.hanazaki542.workers.dev and vanexa-ai-proxy.hanazaki542.workers.dev
- Rotate any credentials or API keys that may have been exposed through the malicious relay
- Use npm audit to identify any dependencies on this package and remove them
- Monitor for similar packages from the same author
Sources
- GitHub Advisory GHSA-f255-3246-f22g · GitHub Advisory Database
Cite this entry
"Malicious code in @ikbal_fadilah_vanexa01/vanexa-agent (npm)." supplychainattack.org, Supply Chain Attack Incident Catalog. Disclosed August 5, 2026; last updated August 5, 2026. https://supplychainattack.org/incident/malicious-code-in-ikbal-fadilah-vanexa01-vanexa-agent-npm-zogcwj
Suggest a correction
Found an error or have a newer source? Corrections to factual errors take priority over new entries.
Related incidents
- resolvedcritical
Malicious code in zyr-agent (npm)
zyr-agent (npm) shipped with malicious code that enables remote command execution through a hardcoded preview-slug endpoint controlled by the package author. The AI agent auto-executes tool calls (including bash commands) returned by the remote endpoint without user confirmation.
npmAI agents & skillsMalicious commitCompromised package - resolvedcritical
Malicious code in @aster110/cc2wechat (npm)
The npm package @aster110/cc2wechat contains malicious code that establishes a remote shell interface via WeChat messages. When the daemon is started, it polls Tencent's iLink Bot API and forwards incoming WeChat messages to locally-hosted AI agents (Claude or Codex) configured with safety gates disabled, allowing arbitrary shell and filesystem operations.
npmAI agents & skillsCompromised package - resolvedcritical
Malicious code in mangomind-agent (npm)
The npm package mangomind-agent contained malicious code that established a persistent WebSocket connection to a hardcoded relay server, enabling remote code execution and AI-agent-driven filesystem access on any host running the package. The vulnerability was discovered and reported by the OpenSSF.
npmAI agents & skillsCompromised packageMalicious commit - resolvedcritical
Malicious code in @lyxa.ai/core (npm)
The npm package @lyxa.ai/core contained malicious code that unconditionally routes all application events through author-controlled cloud infrastructure (CloudAMQP, Redis Cloud, GCP) using embedded credentials, and ships live private keys for GCP and Firebase services, allowing the author to intercept, modify, and trigger arbitrary handlers in any installer's process.
npmAI agents & skillsCompromised packageMalicious maintainer