An active supply chain worm campaign, dubbed SANDWORM_MODE, is spreading through typosquatting and AI toolchain poisoning across at least 19 malicious npm packages. The worm exhibits Shai-Hulud characteristics, incorporating GitHub API exfiltration with DNS fallback, hook-based persistence, SSH propagation, and MCP server injection targeting AI coding assistants. It harvests credentials from developer and CI environments, exfiltrates data via multiple channels, and uses stolen identities to propagate. The campaign also includes a weaponized GitHub Action for CI secret harvesting. The worm employs a multi-stage design with obfuscated loaders, time-gated execution, and extensive configuration options. It targets high-traffic developer utilities, crypto tooling, and AI coding tools, posing a significant threat to the software supply chain.
eBay API MCP Server is an open source local MCP server providing AI assistants with comprehensive access to eBay's Sell APIs. All versions are vulnerable to Environment Variable Injection through the updateEnvFile function. The ebay_set_user_tokens tool allows updating the .env file with new tokens. The updateEnvFile function in src/auth/oauth.ts blindly appends or replaces values without validating them for newlines or quotes. This allows an attacker to inject arbitrary environment variables into the configuration file. An attacker can inject arbitrary environment variables into the .env file. This could lead to configuration overwrites, Denial of Service, and potential RCE. There was no fix for this issue at the time of publication.
Summary
A Reflected Cross-Site Scripting (XSS) vulnerability was discovered in the AI Playground's OAuth callback handler. The `error_description` query parameter was directly interpolated into an HTML script tag without proper escaping, allowing attackers to execute arbitrary JavaScript in the context of the victim's session.
Root cause
The OAuth callback handler in `site/ai-playground/src/server.ts` directly interpolated the `authError` value, sourced from the `error_description` query parameter, into an inline `<script>` tag.
Impact
An attacker could craft a malicious link that, when clicked by a victim, would:
* Steal user chat message history - Access all LLM interactions stored in the user's session.
* Access connected MCP Servers - Interact with any MCP servers connected to the victim's session (public or authenticated/private), potentially allowing the attacker to perform actions on the victim's behalf
Mitigation:
* PR: https://github.com/cloudflare/agents/pull/841 https://github.com/cloudflare/agents/pull/841
* Agents-sdk users should upgrade to agents@0.3.10
* Developers using configureOAuthCallback with custom error handling in their own applications should ensure all user-controlled input is escaped before interpolation.
sf-mcp-server is an implementation of Salesforce MCP server for Claude for Desktop. A command injection vulnerability exists in sf-mcp-server due to unsafe use of child_process.exec when constructing Salesforce CLI commands with user-controlled input. Successful exploitation allows attackers to execute arbitrary shell commands with the privileges of the MCP server process.
CVE-2026-25905 is a medium severity vulnerability affecting the 'mcp-run-python' project, where Python code executed via 'runPython' or 'runPythonAsync' is not properly isolated from JavaScript code. This improper compartmentalization (CWE-653) allows Python code to manipulate the JavaScript environment through Pyodide APIs, potentially enabling an attacker to hijack the MCP server and perform malicious activities such as tool shadowing. The vulnerability has a CVSS score of 5.8, indicating moderate risk, with network attack vector, high attack complexity, no privileges required, and user interaction needed. The project is archived and unlikely to receive patches, increasing the risk for users still relying on it. There are no known exploits in the wild as of now. European organizations using this tool or similar environments integrating Python and JavaScript without proper isolation are at risk, especially those in critical infrastructure or software development sectors. Mitigation requires isolating Python execution environments, restricting Pyodide API access, or migrating away from the unmaintained project. Countries with strong software development ecosystems and critical infrastructure relying on such tools, including Germany, France, the UK, and the Netherlands, are more likely to be affected.
CVE-2026-25904 is a Server-Side Request Forgery (SSRF) vulnerability in the Pydantic-AI MCP Run Python tool, which uses an overly permissive Deno sandbox configuration. This misconfiguration allows Python code executed within the sandbox to access the localhost interface of the host system, potentially enabling SSRF attacks. The project is archived and unlikely to receive a fix, increasing risk for users who continue to rely on it. The vulnerability has a CVSS score of 5.8 (medium severity), reflecting moderate impact and exploitation complexity. Exploitation requires user interaction but no privileges or authentication. While no known exploits are currently in the wild, the vulnerability could allow attackers to access internal services, leading to information disclosure and limited integrity and availability impacts. European organizations using this tool or similar configurations should be cautious, especially those in countries with strong adoption of Python and Deno-based development environments. Mitigation involves discontinuing use of the vulnerable tool, isolating sandbox environments, and restricting network access from sandboxed code. Countries with significant tech sectors and software development communities, such as Germany, France, the UK, and the Netherlands, are most likely to be affected.
A vulnerability was found in r-huijts xcode-mcp-server up to f3419f00117aa9949e326f78cc940166c88f18cb. This affects the function registerXcodeTools of the file src/tools/xcode/index.ts of the component run_lldb. The manipulation of the argument args results in command injection. It is possible to launch the attack remotely. The exploit has been made public and could be used. This product takes the approach of rolling releases to provide continious delivery. Therefore, version details for affected and updated releases are not available. The patch is identified as 11f8d6bacadd153beee649f92a78a9dad761f56f. Applying a patch is advised to resolve this issue.
A vulnerability was identified in XixianLiang HarmonyOS-mcp-server 0.1.0. This vulnerability affects the function input_text. The manipulation of the argument text leads to os command injection. Remote exploitation of the attack is possible. The exploit is publicly available and might be used.
CVE-2026-25546 is a high-severity OS command injection vulnerability in godot-mcp versions prior to 0.1.1. The vulnerability arises because user-controlled input, such as projectPath, is passed unsanitized to a shell execution function, allowing attackers to inject arbitrary shell commands. Exploitation requires local access with limited privileges and some user interaction but can lead to full compromise of the MCP server process. This affects any functionality accepting projectPath, including create_scene, add_node, and load_sprite commands. The vulnerability has been patched in version 0.1.1, and no known exploits are currently reported in the wild. European organizations using godot-mcp in development or production environments should update immediately and restrict access to the MCP server.
MCP TypeScript SDK is the official TypeScript SDK for Model Context Protocol servers and clients. From version 1.10.0 to 1.25.3, cross-client response data leak when a single McpServer/Server and transport instance is reused across multiple client connections, most commonly in stateless StreamableHTTPServerTransport deployments. This issue has been patched in version 1.26.0.
Satellite provides a container image that you can use to run an MCP server locally. The MCP server for Satellite is designed for advanced reporting and data analysis that leverages AI capabilities. You can use it to generate dynamic and comprehensive reports from your Satellite inventory.
The AI Engine plugin for WordPress is vulnerable to Server-Side Request Forgery in all versions up to, and including, 3.3.2 via the 'get_audio' function. This makes it possible for authenticated attackers, with Subscriber-level access and above, to make web requests to arbitrary locations originating from the web application and can be used to query and modify information from internal services, if "Public API" is enabled in the plugin settings, and 'allow_url_fopen' is set to 'On' on the server.
CVE-2025-15061 is a critical OS command injection vulnerability in the Framelink Figma MCP Server's fetchWithRetry method. It allows remote attackers to execute arbitrary code without authentication by exploiting improper validation of user input used in system calls. The flaw enables full compromise of the affected server with high impact on confidentiality, integrity, and availability. No user interaction is required, and the vulnerability has a CVSS score of 9.8, indicating critical severity. Although no known exploits are currently reported in the wild, the ease of exploitation and potential damage make timely mitigation essential. European organizations using this product should prioritize patching or applying mitigations immediately. Countries with significant digital design and creative industries, as well as those with higher adoption of Framelink products, are at greater risk. Defenders must implement strict input validation, network segmentation, and monitor for suspicious activity related to this vulnerability.
CVE-2025-62706 is a denial-of-service vulnerability in Authlib's JSON Web Encryption (JWE) handling when using zip=DEF compression. Versions prior to 1.6.5 perform unbounded DEFLATE decompression without limiting output size, allowing an attacker to craft compressed tokens that expand to arbitrarily large sizes, exhausting system resources. This affects Red Hat Satellite versions >=2.1.0 and <2.9.0. A fix is available in version 2.9.0.
orval generates type-safe JS clients (TypeScript) from any valid OpenAPI v3 or Swagger v2 specification. Prior to 7.18.0, the MCP server generation logic relies on string manipulation that incorporates the summary field from the OpenAPI specification without proper validation or escaping. This allows an attacker to "break out" of the string literal and inject arbitrary code. This vulnerability is fixed in 7.18.0.
A path traversal vulnerability exists in Zen MCP Server before 9.8.2 that allows authenticated attackers to read arbitrary files on the system. The vulnerability is caused by flawed logic in the is_dangerous_path() validation function that uses exact string matching against a blacklist of system directories. Attackers can bypass these restrictions by accessing subdirectories of blacklisted paths.
WeKnora is an LLM-powered framework designed for deep document understanding and semantic retrieval. Prior to version 0.2.5, there is a command injection vulnerability that allows authenticated users to inject stdio_config.command/args into MCP stdio settings, causing the server to execute subprocesses using these injected values. This issue has been patched in version 0.2.5.
Microsoft Playwright MCP Server versions prior to 0.0.40 fails to validate the Origin header on incoming connections. This allows an attacker to perform a DNS rebinding attack via a victim’s web browser and send unauthorized requests to a locally running MCP server, resulting in unintended invocation of MCP tool endpoints.
@sylphxltd/filesystem-mcp v0.5.8 is an MCP server that provides file content reading functionality. Version 0.5.8 of filesystem-mcp contains a critical path traversal vulnerability in its "read_content" tool. This vulnerability arises from improper symlink handling in the path validation mechanism: the resolvePath function checks path validity before resolving symlinks, while fs.readFile resolves symlinks automatically during file access. This allows attackers to bypass directory restrictions by leveraging symlinks within the allowed directory that point to external files, enabling unauthorized access to files outside the intended operational scope.
Dive is an open-source MCP Host Desktop Application that enables integration with function-calling LLMs. A critical Stored Cross-Site Scripting (XSS) vulnerability exists in versions prior to 0.11.1 in the Mermaid diagram rendering component. The application allows the execution of arbitrary JavaScript via `javascript:`. An attacker can exploit this to inject a malicious Model Context Protocol (MCP) server configuration, leading to Remote Code Execution (RCE) on the victim's machine when the node is clicked. Version 0.11.1 fixes the issue.