The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documents the use of multiple third-party libraries for visualization, including Chart.js, ECharts, D3.js, Leaflet, and Three.js. These are fetched from well-known and trusted CDNs including cdnjs.cloudflare.com, cdn.jsdelivr.net, unpkg.com, and esm.sh.
[REMOTE_CODE_EXECUTION]: The skill generates interactive HTML widgets (diagram-html) that execute in a sandboxed iframe. This is the intended architectural design for the platform's visualization capabilities. The instructions include security best practices, such as warning against passing untrusted user input to math.evaluate() and using onerror handlers for script loading.
[DATA_EXFILTRATION]: The skill mentions platform restrictions on network calls (connect-src 'none') within widgets, which serves as an inherent security control against unauthorized data exfiltration. It instructs the agent to pre-fetch and inline data rather than making external calls from the client-side widget.
[COMMAND_EXECUTION]: The skill utilizes a sendPrompt function to interact with the chat interface. This is a platform-provided feature that is rate-limited and requires user interaction to execute, mitigating the risk of automated command injection.

diagram