The Agent Skills Directory

COMMAND_EXECUTION (HIGH): The skill defines a fallback strategy utilizing the Azure CLI (az rest and az kusto). The construction of shell commands involving <kql-query> and <database> parameters is highly susceptible to command injection if the input contains shell metacharacters or nested quotes, particularly when the agent retrieves this input from previous tool outputs.
PROMPT_INJECTION (HIGH): Category 8: Indirect Prompt Injection. The skill is designed to ingest and analyze untrusted external data (logs, telemetry, and IoT data) via kusto_query (SKILL.md).
Ingestion points: Data enters the context through kusto_query and kusto_table_schema_get.
Boundary markers: None. There are no instructions to the agent to treat retrieved log data as data rather than instructions.
Capability inventory: The skill has high-privilege capabilities including KQL execution and Shell command execution through the Azure CLI fallback.
Sanitization: Absent. There is no evidence of escaping or validating the content of retrieved logs before they are used in subsequent reasoning or command construction.
EXTERNAL_DOWNLOADS (LOW): The skill makes network requests to Azure endpoints (e.g., kusto.windows.net) via REST API calls and CLI commands. While targeting a reputable provider, these represent an external data flow that could be used for data exfiltration if the target URL is manipulated.

azure-kusto