Qbox Lua - CfxLua & FiveM

AI Security Review

Evidence context: threat category none; evidence quality strong.

The qbx-lua extension provides CfxLua language support for FiveM resource development, offering native completion, diagnostics, formatting, and navigation. These capabilities inherently require reading and writing files within the active workspace to parse Lua scripts, apply formatting rules, and surface diagnostic information. This level of filesystem access is standard and fully justified for any language server or code intelligence tool. The extension declares no special permissions or host permissions in its manifest, meaning it relies on the default workspace access granted to all Visual Studio Code extensions rather than requesting elevated system-level capabilities.

When examining the security findings, the only flagged item is a medium-severity network detection titled NET-JQUERY_AJAX-extension/dist/assistantMcp.js-1, located in the bundled file extension/dist/assistantMcp.js. This finding triggers because the bundled JavaScript contains jQuery AJAX patterns, which are commonly used to make HTTP requests. In the context of a modern IDE extension, bundled distribution files routinely include networking libraries to fetch completion data, synchronize with remote language servers, or power auxiliary features like an AI assistant. The extension does not declare any specific network endpoints, and no suspicious domains or data exfiltration targets were identified in the bundle. The network activity is isolated to generic library code rather than custom malicious payloads. Automated scanners frequently flag these generic networking patterns in minified bundles, creating noise that does not reflect actual malicious behavior.

There are no findings related to credential access, secret scanning, or postinstall execution. The extension does not attempt to read .env files, SSH keys, or cloud credentials, nor does it execute arbitrary shell commands during activation. The absence of these high-risk patterns confirms that the extension is not attempting to steal developer secrets or establish persistence.

The strongest counterargument to a benign verdict is the presence of assistantMcp.js, which implies the extension includes a Model Context Protocol integration or AI assistant feature. One might argue that an AI assistant could exfiltrate source code to a remote server. However, this feature is explicitly part of the extension functionality for FiveM development, and the network finding itself is merely a generic jQuery AJAX match in a minified bundle. Without evidence of actual exfiltration endpoints or unauthorized data collection, the inclusion of an AI helper does not constitute a security violation. The findings are entirely consistent with normal bundled dependencies and expected IDE behavior.

Key Reasons

• Single network finding is a generic jQuery AJAX match in a bundled dist file
• No malicious indicators, credential access, or suspicious endpoints
• Extension behavior aligns with standard language server capabilities
• Empty permissions and host_permissions arrays indicate no excessive capability requests

False Positive Considerations

• Bundled dependencies in dist/ files triggering network rules
• Generic jQuery AJAX network finding in minified assistantMcp.js

Reviewed 2026-10-09; recommended action: no action; model confidence 90%.