lakehouse-table-inspection
Pass
Audited by Gen Agent Trust Hub on Aug 2, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted metadata from external data catalogs (Unity Catalog, AWS Glue, Polaris), creating a surface for indirect prompt injection. Malicious instructions embedded within table names, column descriptions, or schema metadata could potentially influence the agent's behavior during report generation.
- Ingestion points: Metadata retrieved from external catalogs via
pyiceberganddatabricks.sdk(SKILL.md). - Boundary markers: None present in the instructions to separate external data from agent instructions.
- Capability inventory: The skill produces structured Markdown metadata summaries intended for consumption by humans or other agents.
- Sanitization: No explicit sanitization or validation of the retrieved metadata is described before output generation.
Audit Metadata