The Agent Skills Directory

Official Microsoft Service Integration: The skill connects to a hosted service at https://workiq.svc.cloud.microsoft/mcp, which is an official Microsoft domain. Both the author and the service infrastructure are verified and consistent with the skill's purpose as a first-party workplace intelligence tool.
Secure Credential Management: The instructions emphasize that authentication tokens should be managed by the host environment and specifically warn against including sensitive credentials or tokens in prompts or tool arguments.
User Confirmation for Write Operations: A key safety feature is the requirement for the agent to summarize and get explicit user confirmation before executing any write operations, such as sending emails, deleting events, or reacting to messages. This ensures human oversight for all persistent changes.
Indirect Prompt Injection Surface: Since the skill processes organizational data like emails and messages, it is inherently exposed to potential indirect prompt injection. The skill provides mitigation by instructing the agent to ground responses strictly in tool results and maintain a structured 'resolve-then-act' workflow.
Restrictions on Binary Data: The documentation clarifies that the skill currently does not support the transmission or retrieval of raw binary content, such as file bytes or attachments. Instead, agents are instructed to provide web URLs to the user, which prevents potential exfiltration via raw byte streams.
Transparent Error Reporting: The skill instructs the agent to report failures exactly as they occur and avoid inventing or guessing status codes (such as HTTP 403 or AccessDenied) that were not explicitly observed, which ensures the user is accurately informed about service status.

workiq-preview