The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for utilizing the bq command-line tool and gcloud SDK to manage BigQuery datasets, tables, and jobs. This includes operations like enabling APIs, creating resources, and running SQL queries.
[EXTERNAL_DOWNLOADS]: The skill guides users to install official Google Cloud client libraries for Python, Node.js, and Go, as well as the bigframes library for data analysis. These downloads target official package registries and repositories associated with a trusted organization.
[DYNAMIC_EXECUTION]: The instructions describe the BigQuery MCP server's execute_sql tool, which executes SQL statements at runtime. It also notes that these queries can involve side effects through remote functions or Python User-Defined Functions (UDFs).
[INDIRECT_PROMPT_INJECTION]: The skill interacts with external data sources and processes SQL queries based on user intent, creating a surface for indirect prompt injection.
Ingestion points: Data ingested through bq load commands and user-supplied SQL queries processed by the MCP server or client libraries.
Boundary markers: The skill relies on standard SQL syntax structure; no specific additional boundary markers or "ignore" instructions for embedded data are provided in the skill body.
Capability inventory: The skill possesses capabilities for data retrieval, modification, schema management, and infrastructure deployment via bq, Terraform, and client libraries.
Sanitization: The referenced execute_sql tool in the MCP server is restricted to SELECT statements to mitigate risks associated with unauthorized data modification.

bigquery-basics