answering-natural-language-questions-with-dbt
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides instructions for using the 'jq' command-line utility to filter and extract information from dbt project metadata files such as 'manifest.json' and 'catalog.json'.\n- [DYNAMIC_EXECUTION]: The skill generates SQL queries at runtime based on user questions and discovered schema information, which are then executed through database tools like 'execute_sql'.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user questions and internal project metadata files, creating an ingestion surface for indirect prompt injection.\n
- Ingestion points: User business questions and project artifact files ('target/manifest.json', 'target/catalog.json').\n
- Boundary markers: The instructions do not specify the use of delimiters or 'ignore' warnings for the data ingested from the manifest files.\n
- Capability inventory: The skill utilizes capabilities to execute SQL ('execute_sql'), display query results ('show'), and run shell commands ('jq').\n
- Sanitization: There is no explicit requirement for sanitizing the contents of the dbt manifest or catalog files before the agent processes them.
Audit Metadata