bigquery-graph-author
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external documents (ERDs, ontologies) to extract graph structures, creating a surface for indirect prompt injection. \n
- Ingestion points: User documents and diagrams processed in the Discovery station (SKILL.md, references/document-source.md).\n
- Boundary markers: Instructions in references/document-source.md direct the agent to treat document content as data and ignore instruction-like sentences.\n
- Capability inventory: Execution of SQL and CREATE PROPERTY GRAPH DDL in BigQuery.\n
- Sanitization: Multi-step verification, readable plan approval, and mandatory user confirmation before DDL execution.\n- [COMMAND_EXECUTION]: The skill generates and executes SQL queries against the BigQuery platform to discover metadata and land graph structures. All operations are subject to user approval as specified in the journey stations.
Audit Metadata