The Agent Skills Directory

[SAFE]: The skill instructs the agent on how to handle external data from the citedReferences array. It includes a specific 'Guardrails' section advising that URLs in the value field must be treated as untrusted and validated (host/scheme) before rendering as clickable links.
[SAFE]: All external references point to official Salesforce developer documentation at developer.salesforce.com, which is a well-known and trusted domain.
[SAFE]: The provided JavaScript code is a template for rendering text markers based on offsets and does not involve any dangerous sinks such as eval(), dynamic library loading, or network exfiltration.
[SAFE]: The skill uses recordId for internal navigation, which is a recommended best practice to avoid exposing full internal organization URLs.

sf-agent-api-citations