CVE-2026-134755 in better-chatbot
Summary
by MITRE • 11/22/2026
better-chatbot before 2.0.0 contains a missing authorization vulnerability that lets built-in tools run without user confirmation in Manual tool mode, because src/app/api/chat/route.ts keeps their execute handlers. Attackers can plant prompt-injected documents or web pages that make the HTTP request tool send conversation data to attacker-controlled URLs.
If you want to get the best quality for vulnerability data then you always have to consider VulDB.
Analysis
by VulDB Data Team • 11/22/2026
The identified vulnerability resides within versions of better-chatbot prior to release 2.0.0 and represents a critical failure in access control mechanisms, specifically categorized under CWE-862 Missing Authorization. This flaw allows built-in tools to execute without requiring explicit user confirmation when the application is operating in Manual tool mode. The root cause lies in the implementation details of the server-side route handler located at src/app/api/chat/route.ts, which retains executable handlers for these internal utilities even when such execution should be gated by user consent. This architectural oversight effectively bypasses intended security boundaries, granting unauthorized operational capabilities to components that are not designed to act autonomously without explicit permission in this specific mode of operation.
The technical exploitation vector leverages the interaction between prompt injection and the unauthenticated tool execution capability. An attacker can craft malicious inputs containing prompt-injected documents or web pages designed to manipulate the language model into generating commands for built-in tools, particularly those capable of making external network requests such as HTTP request utilities. Because the authorization check is missing, these injected instructions are processed and executed directly by the backend service without any intermediate validation step that would normally require user approval. This allows the system to act on malicious directives embedded within seemingly benign content, turning a standard conversational interface into an active proxy for data exfiltration or further network reconnaissance.
The operational impact of this vulnerability is severe, primarily centering on confidentiality breaches and potential compromise of downstream systems. By forcing the HTTP request tool to send conversation data to attacker-controlled URLs, adversaries can extract sensitive information processed during the chat session, including personal identifiable information, proprietary business logic, or credentials inadvertently shared by users. Furthermore, if the underlying infrastructure allows outbound connections to arbitrary destinations, this vulnerability could facilitate broader attacks such as server-side request forgery against internal services that trust requests originating from the application host. This aligns with ATT&CK technique T1059 Command and Scripting Interpreter for automated execution of malicious commands and potentially T1048 Exfiltration Over Alternative Protocol if data is sent via non-standard channels, although HTTP exfiltration falls under standard web-based data theft patterns often associated with CWE-200 Information Exposure.
Mitigation strategies must focus on enforcing strict authorization checks before any tool execution occurs in Manual mode. Developers should modify the src/app/api/chat/route.ts file to ensure that all built-in tools require explicit user confirmation or a verified administrative override before their execute handlers are invoked. Implementing an allowlist of permitted actions and requiring cryptographic signatures for tool invocations can further harden this control. Additionally, input validation on documents and web pages should be enhanced to detect and neutralize prompt injection attempts, although defense in depth is essential since no single measure is sufficient against sophisticated adversarial inputs. Upgrading to version 2.0.0 or later resolves the issue by addressing these authorization gaps, ensuring that user intent remains the primary gatekeeper for automated actions within the chatbot environment.