duckdb
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of external datasets from local files and remote S3 buckets, creating a potential surface for instructions embedded within the data to influence agent behavior. * Ingestion points: Local file paths (e.g., sales.csv, data/sales_*.parquet) and remote S3 URIs (e.g., s3://my-bucket/data.parquet) processed via read_parquet or read_csv_auto in SKILL.md. * Capability inventory: SQL execution, network access via the httpfs extension, and local filesystem writes (e.g., analytics.duckdb) described in SKILL.md. * Boundary markers: No explicit prompt delimiters or warnings for external data are provided in the instructions. * Sanitization: No data validation or sanitization logic is described.
- [EXTERNAL_DOWNLOADS]: The skill utilizes the DuckDB extension manager to download and load binary modules at runtime to expand functionality. * Evidence: Usage of 'INSTALL httpfs' and 'LOAD httpfs' commands in SKILL.md to enable network-based file access and remote storage integration.
Audit Metadata