devextreme-chat

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). Runtime path: onMessageEntered (e.message.text from the chat UI) is captured as free-text and then passed into AI calls / rendered back into items (including via messageTemplate HTML rendering), meaning outsider-authored user input can reach the LLM/HTML handling without any prior selection.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The jQuery example imports and executes a remote JS module at runtime from https://esm.sh/micromark@3?bundle, which fetches and runs external code the skill relies on for Markdown rendering.