The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from database system logs which can include untrusted content from user-submitted queries.
Ingestion points: Reads from system.query_log, system.text_log, and system.kafka_consumers in checks.sql and SKILL.md.
Boundary markers: No explicit boundary markers or instructions to ignore embedded content are used when presenting log data to the agent.
Capability inventory: The agent executes SQL queries on ClickHouse system tables to extract performance metrics.
Sanitization: The skill does not sanitize or filter the content of log messages or query strings found in the system tables.
Context: This behavior is standard for database performance analysis skills and the risk is considered low within the intended use-case of monitoring specific database engine behavior.
[COMMAND_EXECUTION]: The skill uses structured SQL queries to inspect the state of the ClickHouse database. The queries are restricted to the system database and focus on performance metrics, such as part creation rates, merge balance, and Kafka consumer status. No dangerous administrative commands or shell executions are present.

altinity-expert-clickhouse-ingestion