@lobehub/lobehub@2.1.9

@lobehub/lobehub is a LobeHub - an open-source,comprehensive AI Agent framework that supports speech synthesis, multimodal, and extensible Function Call plugin system. Supports one-click free deployment of your private ChatGPT/LLM web application.

Affected versions of this package are vulnerable to Authorization Bypass Through User-Controlled Key in the semanticSearch process. An attacker can access other users' text content, file names, and metadata by supplying arbitrary file or knowledge-base identifiers through the chunk retrieval and chat knowledge-base paths.

A fix was pushed into the master branch but not yet published.

@lobehub/lobehub is a LobeHub - an open-source,comprehensive AI Agent framework that supports speech synthesis, multimodal, and extensible Function Call plugin system. Supports one-click free deployment of your private ChatGPT/LLM web application.

Affected versions of this package are vulnerable to Server-side Request Forgery (SSRF) via the importFromUrl and fetchImageFromUrl endpoints. An attacker can access internal network resources and retrieve sensitive information by supplying crafted URLs to these endpoints, which use the global fetch method without proper request filtering.

A fix was pushed into the master branch but not yet published.

@lobehub/lobehub is a LobeHub - an open-source,comprehensive AI Agent framework that supports speech synthesis, multimodal, and extensible Function Call plugin system. Supports one-click free deployment of your private ChatGPT/LLM web application.

Affected versions of this package are vulnerable to Authorization Bypass Through User-Controlled Key through the omission of user identifier checks in the updateMessagePlugin, updatePluginState, updatePluginError, updateTTS, updateTranslate, and findMessagePlugin methods. An attacker can modify another user's plugin tool-call metadata, plugin state, error information, text-to-speech, and translation records by supplying a known message identifier belonging to the victim. This is only exploitable if the attacker knows the victim's non-enumerable message identifier.

A fix was pushed into the master branch but not yet published.

@lobehub/lobehub is a LobeHub - an open-source,comprehensive AI Agent framework that supports speech synthesis, multimodal, and extensible Function Call plugin system. Supports one-click free deployment of your private ChatGPT/LLM web application.

Affected versions of this package are vulnerable to Regular Expression Denial of Service (ReDoS) via the findSkillMd function. An attacker can cause the Node.js event loop to become blocked by supplying a specially crafted basePath value containing unescaped regular expression metacharacters in a GitHub repository URL path during skill import. This results in catastrophic backtracking when the pattern is injected into a dynamically constructed regular expression and executed synchronously against archive entries, leading to denial of service for all concurrent users for an extended period.

A fix was pushed into the master branch but not yet published.

@lobehub/lobehub is a LobeHub - an open-source,comprehensive AI Agent framework that supports speech synthesis, multimodal, and extensible Function Call plugin system. Supports one-click free deployment of your private ChatGPT/LLM web application.

Affected versions of this package are vulnerable to Server-side Request Forgery (SSRF) via the POST /webapi/proxy process. An attacker can make arbitrary outbound requests from the server, leak infrastructure details, and inject authentication cookies by supplying crafted URLs and leveraging reflected response headers. This enables the attacker to set known session values for users through CSRF, potentially gaining unauthorized access to user sessions and abusing the server as an anonymous proxy for further attacks.

There is no fixed version for @lobehub/lobehub.

@lobehub/lobehub is a LobeHub - an open-source,comprehensive AI Agent framework that supports speech synthesis, multimodal, and extensible Function Call plugin system. Supports one-click free deployment of your private ChatGPT/LLM web application.

Affected versions of this package are vulnerable to Command Injection in the message rendering process when handling custom tags in the Renderer component. An attacker can execute arbitrary system commands with the privileges of the current user by inducing the application to process malicious content that triggers the exposed runCommand IPC interface via a crafted LLM provider endpoint.

A fix was pushed into the master branch but not yet published.