The Agent Skills Directory

Indirect Prompt Injection Surface: The skill processes external data from Azure environments, such as resource tags and Advisor recommendations. While this is necessary for optimization analysis, it represents a surface where external content is interpolated into the agent's context. The skill mitigates this by instructing the agent to recommend changes only and requiring explicit human approval for any implementation.\n- Dynamic Command Generation: The skill utilizes generate_query and validate_query for Azure Resource Graph operations. This allows for flexible data retrieval. The inclusion of a validation step and the instruction to avoid local parsing of results are positive security practices that reduce the risk of command injection or unsafe data handling.\n- External Tool Fallback: The skill provides guidance for falling back to Azure CLI (az rest) or PowerShell (Invoke-AzRestMethod) if standard MCP tools are unavailable. These tools provide extensive access to Azure environments. The skill addresses this by requiring explicit approval before any destructive or purchasing actions, maintaining a safe operational boundary.

cost-optimization