The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates AI integration through 'AI Columns' and an 'AI Assistant'. These features ingest data from the grid's dataSource and interpolate it into prompts sent to an LLM provider. This constitutes an indirect prompt injection attack surface where malicious data in the source could influence the agent's output. The skill mitigates this by providing specific event handlers (onAIColumnRequestCreating and onAIAssistantRequestCreating) that allow developers to filter or sanitize the data before it is sent to the AI service.
[EXTERNAL_DOWNLOADS]: The documentation references external scripts and libraries required for exporting data and AI integration. These include devextreme-exceljs-fork, jspdf, and file-saver, which are fetched from well-known and reputable CDNs like cdnjs.cloudflare.com and cdn.jsdelivr.net. These are standard dependencies for the component's functionality.
[REMOTE_CODE_EXECUTION]: No unauthorized or dangerous remote code execution patterns were identified. The network requests described (e.g., to a placeholder https://my-backend.com/ai or OData endpoints) are legitimate data-binding and AI-service communication patterns for a UI component.
[CREDENTIALS_UNSAFE]: The skill correctly advises developers to route AI requests through their own backends and explicitly warns against exposing API keys in client-side code.

devextreme-datagrid