FastGPT's JavaScript sandbox worker attempts to block dynamic import() calls using a regex that only matches ASCII whitespace between 'import' and '('. However, JavaScript syntax permits block comments in this position, which the regex does not account for. An attacker can craft a payload like import/**/("child_process") that is syntactically valid but bypasses the regex filter. Since dynamic import() calls are not wrapped by the safeRequire Proxy (which only proxies require calls), this allows loading the child_process module and executing commands via execSync within the sandbox container running as uid=100. This vulnerability enables arbitrary command execution within the sandbox environment. The issue is resolved in FastGPT version 4.15.0-beta1.

An attacker exploiting this vulnerability can execute arbitrary commands inside the sandbox container with uid=100 privileges by bypassing the dynamic import() filter and loading the child_process module. This can lead to unauthorized code execution and potential compromise of the sandbox environment. The CVSS score is 6.3 (medium severity), reflecting network attack vector, low complexity, requiring low privileges, no user interaction, and impacts on confidentiality, integrity, and availability.

This vulnerability is fixed in FastGPT version 4.15.0-beta1. Users should upgrade to version 4.15.0-beta1 or later to remediate this issue. No other official remediation or temporary fixes are documented. Patch status is not explicitly stated beyond the fix in 4.15.0-beta1; therefore, verify the vendor advisory for the latest remediation guidance.