This vulnerability involves improper neutralization of input during web page generation in pgAdmin 4's cloud deployment module. Exception messages from AWS, Azure, and Google SDKs are included without HTML encoding in JSON response fields (info and errormsg). The Cloud Wizard frontend uses html-react-parser to render these messages, allowing attacker-controlled HTML to be injected into the DOM. The primary attack vector is an authenticated user submitting a crafted access_key that triggers an exception containing the payload, which is then rendered and executed in the victim's browser context. Because the injection occurs within pgAdmin's own interface, standard frame-based mitigations like X-Frame-Options and Content-Security-Policy frame-ancestors do not prevent exploitation. Exploitation is primarily self-targeted, but cross-user attacks require a CSRF token. Multiple endpoints exhibiting this pattern were fixed by introducing a shared sanitize_external_text helper that HTML-escapes exception strings and rendering backend messages in plain-text mode on the frontend. The vulnerability affects pgAdmin 4 versions from 6.6 up to but not including 9.16.

The vulnerability allows an authenticated user to inject malicious HTML and JavaScript into their own pgAdmin 4 interface via cloud deployment-related endpoints, leading to self-targeted cross-site scripting. This can result in redirection of the user's pgAdmin tab to attacker-controlled sites. Exploitation against other authenticated users requires an additional CSRF attack with a valid CSRF token. There is no direct impact on confidentiality or availability reported. The CVSS score is 3.5 (low severity), reflecting limited impact and the requirement for authentication and user interaction.

A patch is available that properly HTML-escapes all external SDK exception strings before including them in JSON responses and renders these messages in plain-text mode on the frontend to prevent HTML parsing. Since pgAdmin 4 is a cloud service, the vendor manages remediation for this cloud-hosted service. Users should upgrade to version 9.16 or later where this issue is fixed. Until then, avoid submitting crafted credentials that could trigger this vulnerability. No additional mitigations such as frame-ancestors or X-Frame-Options are effective against this issue due to its rendering context.