The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documents the use of the bq and gcloud command-line tools for standard administrative tasks such as enabling APIs, creating datasets, and executing SQL queries. These operations are essential for BigQuery management and are presented within the context of intended usage.
[EXTERNAL_DOWNLOADS]: The documentation provides instructions for installing official Google Cloud client libraries using standard package managers, including pip install google-cloud-bigquery, npm install @google-cloud/bigquery, and go get cloud.google.com/go/bigquery. These resources are hosted by well-known and reputable services.
[PROMPT_INJECTION]: In references/bigquery_ai_generate.md, the skill describes how to use the AI.GENERATE function to process data stored in BigQuery tables. This involves interpolating table content directly into prompts (e.g., Summarize this article: ' || article_content). This pattern represents an indirect prompt injection surface where untrusted data within a table could attempt to influence the model's behavior.
Ingestion points: BigQuery table columns (e.g., article_content, invoice_text) processed via AI.GENERATE in references/bigquery_ai_generate.md.
Boundary markers: None explicitly shown in the prompt concatenation examples.
Capability inventory: The skill utilizes BigQuery's SQL execution engine and integrated LLM features.
Sanitization: No explicit sanitization or validation of the table data is demonstrated in the provided examples.

bigquery-basics