The Agent Skills Directory

[Data Exposure & Exfiltration] (LOW): The skill identifies and uses standard Azure CLI commands (az monitor activity-log, az containerapp logs show) and MCP tools (mcp_azure_mcp_monitor) to access cloud logs. While these commands handle potentially sensitive diagnostic data, they are standard operational tools for the stated purpose of 'troubleshooting production issues.' No non-whitelisted external network destinations or hardcoded credentials were found.
[Indirect Prompt Injection] (LOW): The skill processes external data (Azure logs and resource health data) via the mcp_azure_mcp_monitor and mcp_azure_mcp_applens tools. There is a low risk of 'Log Injection' where an attacker could place malicious instructions inside application logs to influence the agent's reasoning. However, the skill's capabilities are limited to diagnostic read operations and documentation of findings, which minimizes the risk of side effects.
[Command Execution] (LOW): The skill provides templates for standard Azure CLI commands. These are intended for local execution by the user or an agent with appropriate Azure permissions. The commands are static and do not involve suspicious pipe-to-bash or dynamic execution patterns.

azure-diagnostics