Infostealer Malware Targets OpenClaw Configuration Files

OpenClaw, an AI agent platform, faces significant security concerns as attackers exploit its ecosystem. Malicious skills on ClawHub, a public skills registry, have been discovered, and threat actors are discussing the deployment of OpenClaw skills for botnet operations. The number of malicious packages on npm and PyPI with the name 'claw' has surged, providing new avenues for threat actors. Additionally, attackers are actively scanning exposed OpenClaw gateways, attempting prompt injection and command execution. These developments highlight the risks associated with AI agents' broad permissions and unsupervised deployment.

A security audit by Koi Security identified 341 malicious skills on ClawHub, a marketplace for OpenClaw users, which distribute Atomic Stealer malware to steal sensitive data from macOS and Windows systems. The campaign, codenamed ClawHavoc, uses social engineering tactics to trick users into installing malicious prerequisites. The skills masquerade as legitimate tools, including cryptocurrency utilities, YouTube tools, and finance applications. OpenClaw has added a reporting feature and partnered with VirusTotal to scan skills uploaded to ClawHub, providing an additional layer of security for the OpenClaw community. The malware targets API keys, credentials, and other sensitive data, exploiting the open-source ecosystem's vulnerabilities. The campaign coincides with a report from OpenSourceMalware, highlighting the same threat. The intersection of AI agent capabilities and persistent memory amplifies the risks, enabling stateful, delayed-execution attacks. New findings reveal almost 400 fake crypto trading add-ons in the project behind the viral Moltbot/OpenClaw AI assistant tool can lead users to install information-stealing malware. These addons, called skills, masquerade as cryptocurrency trading automation tools and target ByBit, Polymarket, Axiom, Reddit, and LinkedIn. The malicious skills share the same command-and-control (C2) infrastructure, 91.92.242.30, and use sophisticated social engineering to convince users to execute malicious commands which then steals crypto assets like exchange API keys, wallet private keys, SSH credentials, and browser passwords.

A malicious Microsoft Visual Studio Code (VS Code) extension named "ClawdBot Agent - AI Coding Assistant" was discovered on the official Extension Marketplace. The extension, which posed as a free AI coding assistant, stealthily dropped a malicious payload on compromised hosts. The extension was taken down by Microsoft after being reported by cybersecurity researchers. The malicious extension executed a binary named "Code.exe" that deployed a legitimate remote desktop program, granting attackers persistent remote access to compromised hosts. The extension also incorporated multiple fallback mechanisms to ensure payload delivery, including retrieving a DLL from Dropbox and using hard-coded URLs to obtain the payloads. Additionally, security researchers found hundreds of unauthenticated Moltbot instances online, exposing sensitive data and credentials. Moltbot, an open-source personal AI assistant, can run 24/7 locally, maintaining a persistent memory and executing scheduled tasks. However, insecure deployments can lead to sensitive data leaks, corporate data exposure, credential theft, and command execution. Hundreds of Clawdbot Control admin interfaces are exposed online due to reverse proxy misconfiguration, allowing unauthenticated access and root-level system access. More than 230 malicious packages for OpenClaw (formerly Moltbot and ClawdBot) have been published in less than a week on the tool's official registry and on GitHub. These malicious skills impersonate legitimate utilities and inject information-stealing malware payloads onto users' systems, targeting sensitive data like API keys, wallet private keys, SSH credentials, and browser passwords. Users are advised to audit their configurations, revoke connected service integrations, and implement network controls to mitigate potential risks. A self-styled social networking platform built for AI agents, Moltbook, contained a misconfigured database that allowed full read and write access to all data. The exposure was due to a Supabase API key exposed in client-side JavaScript, granting unauthenticated access to the entire production database. Researchers accessed 1.5 million API authentication tokens, 30,000 email addresses, and thousands of private messages between agents. The API key exposure allowed attackers to impersonate any agent on the platform, post content, send messages, and interact as that agent. Unauthenticated users could edit existing posts, inject malicious content or prompt injection payloads, and deface the site. SecurityScorecard found 40,214 exposed OpenClaw instances associated with 28,663 unique IP addresses. 63% of observed deployments are vulnerable, with 12,812 instances exploitable via remote code execution (RCE) attacks. SecurityScorecard correlated 549 instances with prior breach activity and 1493 with known vulnerabilities. Three high-severity CVEs in OpenClaw have been discovered, with public exploit code available. OpenClaw instances are at risk of indirect prompt injection and API key leaks, with most exposures located in China, the US, and Singapore.

The GlassWorm malware campaign has resurfaced with a third wave, adding 24 new packages to OpenVSX and Microsoft Visual Studio Marketplace. The malware uses invisible Unicode characters to hide malicious code and targets GitHub, NPM, and OpenVSX account credentials, as well as cryptocurrency wallet data. The campaign initially impacted 49 extensions, with an estimated 35,800 downloads, though this figure includes inflated numbers due to bots and visibility-boosting tactics. The Eclipse Foundation has revoked leaked tokens and introduced security measures, but the threat actors have pivoted to GitHub and now returned to OpenVSX with updated command-and-control endpoints. The malware's global reach includes systems in the United States, South America, Europe, Asia, and a government entity in the Middle East. Koi Security has accessed the attackers' server and shared victim data with law enforcement. The threat actors have posted a fresh transaction to the Solana blockchain, providing an updated C2 endpoint for downloading the next-stage payload. The attacker's server was inadvertently exposed, revealing a partial list of victims spanning the U.S., South America, Europe, and Asia, including a major government entity from the Middle East. The threat actor is assessed to be Russian-speaking and uses the open-source browser extension C2 framework named RedExt as part of their infrastructure. The third wave of Glassworm uses Rust-based implants packaged inside the extensions and targets popular tools and developer frameworks like Flutter, Vim, Yaml, Tailwind, Svelte, React Native, and Vue. Additionally, a malicious Rust package named "evm-units" was discovered, targeting Windows, macOS, and Linux systems. This package, uploaded to crates.io in mid-April 2025, attracted over 7,000 downloads and was designed to stealthily execute on developer machines by masquerading as an Ethereum Virtual Machine (EVM) unit helper tool. The package checks for the presence of Qihoo 360 antivirus and alters its execution flow accordingly. The references to EVM and Uniswap indicate that the supply chain incident is designed to target developers in the Web3 space. The latest development involves the compromise of a legitimate developer's resources to push malicious updates to downstream users, with the malicious extensions having previously been presented as legitimate developer utilities and collectively accumulated over 22,000 Open VSX downloads prior to the malicious releases. A new GlassWorm malware attack through compromised OpenVSX extensions focuses on stealing passwords, crypto-wallet data, and developer credentials and configurations from macOS systems. The threat actor gained access to the account of a legitimate developer (oorzc) and pushed malicious updates with the GlassWorm payload to four extensions that had been downloaded 22,000 times. GlassWorm attacks first appeared in late October, hiding the malicious code using "invisible" Unicode characters to steal cryptocurrency wallet and developer account details. The malware also supports VNC-based remote access and SOCKS proxying. Over time and across multiple attack waves, GlassWorm impacted both Microsoft's official Visual Studio Code marketplace and its open-source alternative for unsupported IDEs, OpenVSX. In a previous campaign, GlassWorm showed signs of evolution, targeting macOS systems, and its developers were working to add a replacement mechanism for the Trezor and Ledger apps. A new report from Socket's security team describes a new campaign that relied on trojanizing the following extensions: oorzc.ssh-tools v0.5.1, oorzc.i18n-tools-plus v1.6.8, oorzc.mind-map v1.0.61, oorzc.scss-to-css-compile v1.3.4. The malicious updates were pushed on January 30, and Socket reports that the extensions had been innocuous for two years. This suggests that the oorzc account was most likely compromised by GlassWorm operators. According to the researchers, the campaign targets macOS systems exclusively, pulling instructions from Solana transaction memos. Notably, Russian-locale systems are excluded, which may hint at the origin of the attacker. GlassWorm loads a macOS information stealer that establishes persistence on infected systems via a LaunchAgent, enabling execution at login. It harvests browser data across Firefox and Chromium, wallet extensions and wallet apps, macOS keychain data, Apple Notes databases, Safari cookies, developer secrets, and documents from the local filesystem, and exfiltrates everything to the attacker's infrastructure at 45.32.150[.]251. Socket reported the packages to the Eclipse Foundation, the operator of the Open VSX platform, and the security team confirmed unauthorized publishing access, revoked tokens, and removed the malicious releases. The only exception is oorzc.ssh-tools, which was removed completely from Open VSX due to discovering multiple malicious releases. Currently, versions of the affected extensions on the market are clean, but developers who downloaded the malicious releases should perform a full system clean-up and rotate all their secrets and passwords.

APT36, also known as Transparent Tribe, is targeting both Windows and BOSS Linux systems in ongoing attacks against Indian government and defense entities. The campaign, active since June 2025, involves phishing emails delivering malicious .desktop files disguised as PDFs. The malware facilitates data exfiltration, persistent espionage access, and includes anti-debugging and anti-sandbox checks. The malware also targets the Kavach 2FA solution used by Indian government agencies. The attack leverages the .desktop file's 'Exec=' field to execute a sequence of shell commands that download and run a Go-based ELF payload. The payload establishes persistence through cron jobs and systemd services, and communicates with a C2 server via a WebSocket channel. The technique allows APT36 to evade detection by abusing a legitimate Linux feature that is not typically monitored for threats. The campaign demonstrates APT36's evolving tactics, becoming more evasive and sophisticated. The campaign uses dedicated staging servers for malware distribution, transitioning from cloud storage platforms. The malware includes multiple persistence methods and supports commands for file browsing, collection, and remote execution. The campaign is part of a broader trend of targeted activity by South and East Asian threat actors, reflecting a trend toward purpose-built malware and infrastructure. Indian government entities have been targeted in two campaigns codenamed Gopher Strike and Sheet Attack. Gopher Strike leveraged phishing emails to deliver PDF documents with a blurred image and a fake Adobe Acrobat Reader DC update dialog. The campaign uses server-side checks to prevent automated URL analysis tools from fetching the ISO file, ensuring delivery only to intended targets in India. The malicious payload is a Golang-based downloader called GOGITTER, which creates a VBScript file to fetch commands from C2 servers. GOGITTER sets up persistence using a scheduled task to run the VBScript file every 50 minutes. GOGITTER downloads a ZIP file from a private GitHub repository and executes a lightweight Golang-based backdoor called GITSHELLPAD. GITSHELLPAD polls the C2 server every 15 seconds for commands and supports six different commands including cd, run, upload, and download. The results of command execution are stored in a file called "result.txt" and uploaded to the GitHub account. The threat actor also downloads RAR archives containing utilities to gather system information and drop GOSHELL, a bespoke Golang-based loader. GOSHELL's size was artificially inflated to approximately 1 gigabyte to evade detection by antivirus software. GOSHELL only executes on specific hostnames by comparing the victim's hostname against a hard-coded list. APT36 and SideCopy are launching cross-platform RAT campaigns against Indian entities using malware families like Geta RAT, Ares RAT, and DeskRAT. The campaigns use phishing emails with malicious attachments or download links to deliver the malware, which provides persistent remote access, system reconnaissance, data collection, and command execution. Geta RAT supports various commands including system information collection, process enumeration, credential gathering, and file operations. Ares RAT is a Python-based RAT that can run commands issued by the threat actor. DeskRAT is delivered via a rogue PowerPoint Add-In file with embedded macros. The campaigns target Indian defense, government, and strategic sectors, demonstrating a well-resourced, espionage-focused threat actor deliberately targeting these sectors through defense-themed lures, impersonated official documents, and regionally trusted infrastructure.