query-analytical-engine
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns, obfuscation, or security vulnerabilities were detected in the skill's instructions or reference materials.
- [INDIRECT_PROMPT_INJECTION]: Analyzed the surface area for untrusted data processed from external sources.
- Ingestion points: Remote PostgreSQL/MySQL databases and S3 Parquet/Iceberg files accessed via engine table functions as described in
references/analytical-query-guide.md. - Boundary markers: The skill requires the use of structured output contracts (
data-analysis-report.json) and mandates a specific Python validation wrapper for all queries. - Capability inventory: In-process SQL execution (chDB, DuckDB), file system access for report generation, and network connectivity for database federation.
- Sanitization: Mandatory Abstract Syntax Tree (AST) validation using
sqlglotto enforce read-only execution, reject DDL/DML operations, and limit query complexity (joins and row limits). - [COMMAND_EXECUTION]: The skill executes analytical SQL queries within embedded engines. This is protected by the
readonly = 1setting and pre-execution AST validation to prevent unauthorized operations or data mutation. - [EXTERNAL_DOWNLOADS]: Fetches analytical data from Amazon S3 and remote SQL databases. These operations target well-known cloud services and are fundamental to the skill's intended analytical purpose.
Audit Metadata