The vulnerability exists in the nltk.app.wordnet_app module, specifically in the lookup_ route where the 'word' parameter is derived from attacker-controlled base64-encoded pickle data decoded by Reference.decode(). This 'word' is reflected directly into the HTML response without escaping, unlike the search route which properly escapes input. Because the server listens on all interfaces by default, an attacker can craft a URL that injects arbitrary HTML/JavaScript into the response page, resulting in reflected XSS. The issue was demonstrated with a proof-of-concept payload that injects a script alert.

An attacker who can trick a user into opening a maliciously crafted lookup_ URL can execute arbitrary JavaScript in the browser context of the local WordNet Browser application. This can lead to manipulation of page content, execution of same-origin requests within the app, and potentially triggering UI actions. The impact is limited to users running the local or self-hosted WordNet Browser service and opening attacker-controlled links.

A patch is available for this vulnerability. Users should update to a fixed version of the nltk package that properly escapes user input in the lookup_ route. Until patched, users should avoid opening untrusted lookup_ URLs in the WordNet Browser application.