TUUI is a desktop MCP client designed as a tool unitary utility integration. Prior to 1.3.4, a critical Remote Code Execution (RCE) vulnerability exists in Tuui due to an unsafe Cross-Site Scripting (XSS) flaw in the Markdown rendering component. Tuui allows the execution of arbitrary JavaScript within ECharts code blocks. Combined with an exposed IPC interface that allows spawning processes, an attacker can execute arbitrary system commands on the victim's machine simply by having them view a malicious Markdown message. This vulnerability is fixed in 1.3.4.

CodeWhale is a DeepSeek + MiMo coding agent in terminal. Prior to 0.8.26, the task_create tool spawns durable sub-agents that inherit two insecure defaults, allow_shell defaults to true (config.rs:1499: self.allow_shell.unwrap_or(true)) and auto_approve defaults to true (task_manager.rs:297: auto_approve: Some(true)). When a user approves a task_create call (which requires ApprovalRequirement::Required), they approve what appears to be a benign work prompt. However, the spawned sub-agent silently receives unrestricted, unapproved shell access. This vulnerability is fixed in 0.8.26.

Cherry Studio is a desktop client that supports for multiple LLM providers. From versions 1.4.8 to 1.5.0, there is a one-click remote code execution vulnerability through the custom URL handling. An attacker can exploit this by hosting a malicious website or embedding a specially crafted URL on any website. If a victim clicks the exploit link in their browser, the app’s custom URL handler is triggered, leading to remote code execution on the victim’s machine. This issue has been patched in version 1.5.1.

soundcloud-rpc is a SoundCloud Client with Discord Rich Presence, Dark Mode, Last.fm and AdBlock support. Prior to 0.1.8, a track title containing an HTML payload executed locally in the Electron app. This means attacker-controlled SoundCloud track metadata can lead to local command execution on the user's machine. The application exposes a preload API (window.soundcloudAPI.sendTrackUpdate) to the remote SoundCloud page. Track metadata from SoundCloud is trusted and forwarded through IPC into the Electron main process. The app later renders that metadata as raw HTML inside privileged Electron views that have Node.js integration enabled. This vulnerability is fixed in 0.1.8.

SiYuan is an open-source personal knowledge management system. Prior to 3.7.0, the kernel stores Attribute View (AV / database) names without any HTML escape, then a render template uses raw strings.ReplaceAll(tpl, "${avName}", nodeAvName) to embed the name in HTML before pushing to all clients via WebSocket. Three independent client paths (render.ts:120 → outerHTML, Title.ts:401 → innerHTML, transaction.ts:559 → innerHTML) consume the value without escaping. Because the main BrowserWindow runs nodeIntegration:true, contextIsolation:false, webSecurity:false (app/electron/main.js:407-411), HTML injection in the renderer becomes Node.js code execution. This vulnerability is fixed in 3.7.0.

AIRI is a self-hosted, artificial intelligence based Grok Companion. In v0.7.2-beta.2 in the `packages/stage-ui/src/components/MarkdownRenderer.vue` path, the Markdown content is processed using the useMarkdown composable, and the processed HTML is rendered directly into the DOM using v-html. An attacker creates a card file containing malicious HTML/JavaScript, then simply processes it using the highlightTagToHtml function (which simply replaces template tags without HTML escaping), and then directly renders it using v-html, leading to cross-site scripting (XSS). The project also exposes the Tauri API, which can be called from the frontend. The MCP plugin exposes a command execution interface function in `crates/tauri-plugin-mcp/src/lib.rs`. This allows arbitrary command execution. `connect_server` directly passes the user-supplied `command` and `args` parameters to `Command::new(command).args(args)` without any input validation or whitelisting. Thus, the previous XSS exploit could achieve command execution through this interface. v0.7.2-beta.3 fixes the issue.

Insufficient data validation in Permission Prompts in Google Chrome prior to 117.0.5938.62 allowed an attacker who convinced a user to install a malicious app to potentially perform a sandbox escape via a malicious file. (Chromium security severity: Medium)

DeepChat is an open-source artificial intelligence agent platform that unifies models, tools, and agents. Prior to v1.0.4-beta.1, An incomplete mitigation for CVE-2025-55733 leaves DeepChat vulnerable to an arbitrary protocol execution bypass (RCE). While the patch correctly restricted api.openExternal() inside the renderer's preload/index.ts script, it structurally neglected to sanitize native Electron pop-up window handlers. An attacker or a compromised AI endpoint returning a Markdown link can trigger a target="_blank" native window interception in tabPresenter.ts, which forwards the malicious URL directly to shell.openExternal(url) and completely bypasses the isValidExternalUrl security boundary. This vulnerability is fixed in v1.0.4-beta.1.

Arbitrary Code Execution through Sanitizer Bypass in GitHub repository jgraph/drawio prior to 18.0.0. - Arbitrary (remote) code execution in the desktop app. - Stored XSS in the web app.

PraisonAI is a multi-agent teams system. Prior to version 4.6.34, PraisonAI's MCP (Model Context Protocol) server (praisonai mcp serve) registers four file-handling tools by default — praisonai.rules.create, praisonai.rules.show, praisonai.rules.delete, and praisonai.workflow.show. Each accepts a path or filename string from MCP tools/call arguments and joins it onto ~/.praison/rules/ (or, for workflow.show, accepts an absolute path) with no containment check. The JSON-RPC dispatcher passes params["arguments"] blind to each handler via **kwargs without validating against the advertised input schema. By setting rule_name="../../<some-path>" an attacker walks out of the rules directory and writes any file the running user can write. Dropping a Python .pth file into the user site-packages directory escalates this primitive to arbitrary code execution in any subsequent Python process the user spawns — the next praisonai CLI invocation, an IDE script run, the user's python REPL, or any background Python service. This issue has been patched in version 4.6.34.

Notesnook is a note-taking app focused on user privacy & ease of use. Prior to Notesnook Web/Desktop version 3.3.15 and prior to Notesnook iOS/Android version 3.3.20, a stored XSS vulnerability in the note export flow can be escalated to remote code execution in the desktop app. The root cause is that exported note fields such as title, headline, and content are inserted into the generated HTML template without HTML escaping. When the note is later exported to PDF, Notesnook renders that HTML into a same-origin, unsandboxed iframe using iframe.srcdoc = .... Injected script executes in the Notesnook origin. In the desktop app, this becomes RCE because Electron is configured with nodeIntegration: true and contextIsolation: false. This issue has been patched in Notesnook Web/Desktop version 3.3.15 and Notesnook iOS/Android version 3.3.20.

A vulnerability in the LightGlue model loading path of huggingface/transformers version 5.2.0 allows an attacker-controlled model repository to execute arbitrary code during model initialization. The issue arises because the `trust_remote_code` parameter, intended to prevent remote code execution, is overridden by untrusted serialized configuration data in a nested code path. Specifically, when loading a LightGlue model using `AutoModel.from_pretrained()` with `trust_remote_code=False`, the `LightGlueConfig` reads the `trust_remote_code` value from the untrusted `config.json` file and propagates it into nested `AutoConfig.from_pretrained()` calls. This results in the execution of attacker-provided Python modules, even when the victim explicitly disables remote code execution. The vulnerability poses a high risk for environments such as API inference servers, research notebooks, CI/CD pipelines, and model evaluation workers, potentially leading to credential theft, lateral movement, or persistence/backdoor deployment.

nbgitpuller is a Jupyter server extension to sync a git repository one-way to a local path. Due to unsanitized input, visiting maliciously crafted links could result in arbitrary code execution in the user environment. This has been resolved in version 0.10.2 and all users are advised to upgrade. No work around exist for users who can not upgrade.

PraisonAI is a multi-agent teams system. Prior to 4.5.128, PraisonAI treats remotely fetched template files as trusted executable code without integrity verification, origin validation, or user confirmation, enabling supply chain attacks through malicious templates. This vulnerability is fixed in 4.5.128.

Insufficient data validation in Navigation in Google Chrome prior to 145.0.7632.159 allowed a remote attacker to potentially perform a sandbox escape via a crafted HTML page. (Chromium security severity: High)

Vikunja is an open-source self-hosted task management platform. Starting in version 0.21.0 and prior to version 2.2.0, the Vikunja Desktop Electron wrapper enables `nodeIntegration` in the renderer process without `contextIsolation` or `sandbox`. This means any cross-site scripting (XSS) vulnerability in the Vikunja web frontend -- present or future -- automatically escalates to full remote code execution on the victim's machine, as injected scripts gain access to Node.js APIs. Version 2.2.0 fixes the issue.

Notesnook is a note-taking app. Prior to version 3.3.11 on Web/Desktop and 3.3.17 on Android/iOS, a stored XSS in the Web Clipper rendering flow can be escalated to remote code execution in the desktop app. The root cause is that the clipper preserves attacker-controlled attributes from the source page’s root element and stores them inside web-clip HTML. When the clip is later opened, Notesnook renders that HTML into a same-origin, unsandboxed iframe using `contentDocument.write(...)`. Event-handler attributes such as `onload`, `onclick`, or `onmouseover` execute in the Notesnook origin. In the desktop app, this becomes RCE because Electron is configured with `nodeIntegration: true` and `contextIsolation: false`. Version 3.3.11 Web/Desktop and 3.3.17 on Android/iOS patch the issue.

OpenBao is an open source identity-based secrets management system. Prior to version 2.5.2, OpenBao installations that have an OIDC/JWT authentication method enabled and a role with `callback_mode=direct` configured are vulnerable to XSS via the `error_description` parameter on the page for a failed authentication. This allows an attacker access to the token used in the Web UI by a victim. The `error_description` parameter has been replaced with a static error message in v2.5.2. The vulnerability can be mitigated by removing any roles with `callback_mode` set to `direct`.

Insufficient validation of untrusted input in WebNN in Google Chrome on Windows prior to 149.0.7827.53 allowed a remote attacker who had compromised the renderer process to potentially perform a sandbox escape via a crafted HTML page. (Chromium security severity: Medium)

A flaw was found in the Undertow HTTP server core, which is used in WildFly, JBoss EAP, and other Java applications. The Undertow library fails to properly validate the Host header in incoming HTTP requests.As a result, requests containing malformed or malicious Host headers are processed without rejection, enabling attackers to poison caches, perform internal network scans, or hijack user sessions.

Langroid is a framework for building large-language-model-powered applications. Prior to version 0.59.32, there is a bypass to the fix for CVE-2025-46724. TableChatAgent can call pandas_eval tool to evaluate the expression. There is a WAF in langroid/utils/pandas_utils.py introduced to block code injection CVE-2025-46724. However it can be bypassed due to _literal_ok() returning False instead of raising UnsafeCommandError on invalid input, combined with unrestricted access to dangerous dunder attributes (__init__, __globals__, __builtins__). This allows chaining whitelisted DataFrame methods to leak the eval builtin and execute arbitrary code. This issue has been patched in version 0.59.32.

A DNS rebinding vulnerability in the UPnP IGD implementations in Freebox v5 before 1.5.29 and Freebox Server before 4.2.3.

XWiki Platform is a generic wiki platform offering runtime services for applications built on top of it. In affected versions it's possible to execute a content with the right of any user via a crafted URL. A user must have `programming` privileges in order to exploit this vulnerability. This issue has been patched in XWiki 14.10.7 and 15.2RC1. Users are advised to upgrade. There are no known workarounds for for this vulnerability.

A DNS rebinding vulnerability in Freebox v5 before 1.5.29.

A DNS rebinding vulnerability in the Freebox OS web interface in Freebox Server before 4.2.3.

Cherry Studio is a desktop client that supports for multiple LLM providers. Cherry Studio registers a custom protocol called `cherrystudio://`. When handling the MCP installation URL, it parses the base64-encoded configuration data and directly executes the command within it. In the files `src/main/services/ProtocolClient.ts` and `src/main/services/urlschema/mcp-install.ts`, when receiving a URL of the `cherrystudio://mcp` type, the `handleMcpProtocolUrl` function is called for processing. If an attacker crafts malicious content and posts it on a website or elsewhere (there are many exploitation methods, such as creating a malicious website with a button containing this malicious content), when the user clicks it, since the pop-up window contains normal content, the direct click is considered a scene action, and the malicious command is directly triggered, leading to the user being compromised. As of time of publication, no known patched versions exist.

An improper control of generation of code ('code injection') in Fortinet FortiClientLinux version 7.2.0, 7.0.6 through 7.0.10 and 7.0.3 through 7.0.4 allows attacker to execute unauthorized code or commands via tricking a FortiClientLinux user into visiting a malicious website

com.xwiki.identity-oauth:identity-oauth-ui is a package to aid in building identity and service providers based on OAuth authorizations. When a user logs in via the OAuth method, the identityOAuth parameters sent in the GET request is vulnerable to cross site scripting (XSS) and XWiki syntax injection. This allows remote code execution via the groovy macro and thus affects the confidentiality, integrity and availability of the whole XWiki installation. The issue has been fixed in Identity OAuth version 1.6. There are no known workarounds for this vulnerability and users are advised to upgrade.

A flaw was found in keycloak in versions before 13.0.0. A Self Stored XSS attack vector escalating to a complete account takeover is possible due to user-supplied data fields not being properly encoded and Javascript code being used to process the data. The highest threat from this vulnerability is to data confidentiality and integrity as well as system availability.

FACTION is a PenTesting Report Generation and Collaboration Framework. Prior to version 1.7.1, an extension execution path in Faction’s extension framework permits untrusted extension code to execute arbitrary system commands on the server when a lifecycle hook is invoked, resulting in remote code execution (RCE) on the host running Faction. Due to a missing authentication check on the /portal/AppStoreDashboard endpoint, an attacker can access the extension management UI and upload a malicious extension without any authentication, making this vulnerability exploitable by unauthenticated users. This issue has been patched in version 1.7.1.

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.

esm.sh is a nobuild content delivery network(CDN) for modern web development. Prior to version 136, The esm.sh CDN service contains a Template Literal Injection vulnerability (CWE-94) in its CSS-to-JavaScript module conversion feature. When a CSS file is requested with the ?module query parameter, esm.sh converts it to a JavaScript module by embedding the CSS content directly into a template literal without proper sanitization. An attacker can inject malicious JavaScript code using ${...} expressions within CSS files, which will execute when the module is imported by victim applications. This enables Cross-Site Scripting (XSS) in browsers and Remote Code Execution (RCE) in Electron applications. This issue has been patched in version 136.

The Zoom Client for Meetings (for Android, iOS, Linux, macOS, and Windows) before version 5.11.0 are susceptible to a URL parsing vulnerability. If a malicious Zoom meeting URL is opened, the malicious link may direct the user to connect to an arbitrary network address, leading to additional attacks including the potential for remote code execution through launching executables from arbitrary paths.

A vulnerability has been discovered in the Citrix Secure Access client for Ubuntu which, if exploited, could allow an attacker to remotely execute code if a victim user opens an attacker-crafted link and accepts further prompts.

BinderHub is a kubernetes-based cloud service that allows users to share reproducible interactive computing environments from code repositories. In affected versions a remote code execution vulnerability has been identified in BinderHub, where providing BinderHub with maliciously crafted input could execute code in the BinderHub context, with the potential to egress credentials of the BinderHub deployment, including JupyterHub API tokens, kubernetes service accounts, and docker registry credentials. This may provide the ability to manipulate images and other user created pods in the deployment, with the potential to escalate to the host depending on the underlying kubernetes configuration. Users are advised to update to version 0.2.0-n653. If users are unable to update they may disable the git repo provider by specifying the `BinderHub.repo_providers` as a workaround.