The Agent Skills Directory

[Command Execution]: The skill's fallback strategy involves executing Azure CLI (az) commands and REST API calls via the shell. While these are standard developer tools for managing Azure resources, running shell commands involves a higher degree of capability that should be monitored.
Evidence: The skill provides templates for az kusto cluster list and az rest commands to be executed if standard tools fail.
[Indirect Prompt Injection]: The skill is designed to query and process data from external Kusto databases. If a database table contains malicious instructions or specially crafted content, the agent could potentially process these as instructions rather than data.
Ingestion points: Data enters the agent context through the kusto_query tool and the results of az rest calls.
Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are defined for the query results.
Capability inventory: The agent has the capability to execute shell commands (via the Azure CLI fallback) and perform network requests (via the REST API fallback).
Sanitization: There is no evidence of sanitization or filtering applied to the query results before they are processed by the agent.

azure-kusto