The Agent Skills Directory

Read-Only Safety Enforcement: The skill implements a robust safety model by explicitly instructing the agent to only execute read-only PowerShell commands such as Get-*, Test-*, and Export-*. This prevents the agent from making unauthorized changes to the tenant configuration.
Human-in-the-Loop Remediation: Any mutating commands (e.g., Set-*, New-*, Remove-*) are categorized as recommendations only. The instructions require the agent to present these commands as text for an administrator to review and execute manually, ensuring human oversight for all corrective actions.
Structured Diagnostic Pathing: Troubleshooting is guided by symptom-specific decision trees and detailed reference files. This structured approach ensures the agent follows established diagnostic procedures for complex scenarios like retention conflicts, archive provisioning, and policy distribution errors.
Official Resource Utilization: The skill utilizes standard Microsoft 365 and Exchange Online PowerShell cmdlets. External resources and useful articles link directly to official Microsoft documentation (learn.microsoft.com) and trusted diagnostic short-links (aka.ms).
Comprehensive Audit Trail: The workflow includes the use of an execution log tool to provide a full audit trail of all diagnostic commands run, which is a security best practice for administrative agents.

dlm-diagnostics