The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the bq CLI tool to perform data exploration tasks such as listing tables, showing schemas, and querying row counts. These operations are essential for the skill's stated purpose.
[COMMAND_EXECUTION]: A localized python3 script is employed to process and filter JSON schema data. The script performs basic string manipulation and does not execute arbitrary code or access external networks.
[PROMPT_INJECTION]: The skill processes external data from BigQuery, which constitutes a surface for indirect prompt injection.
Ingestion points: BigQuery table names, schemas, and sample row data retrieved via bq commands in SKILL.md.
Boundary markers: Output from database commands is not explicitly wrapped in boundary markers.
Capability inventory: The skill utilizes bq for shell execution and python3 for data processing.
Sanitization: SQL queries utilize backtick escaping for identifiers like project and dataset names, providing standard protection against schema-based injection.

explore-lakehouse