skills/awslabs/agent-plugins/dsql/Gen Agent Trust Hub

dsql

Pass

Audited by Gen Agent Trust Hub on Oct 10, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads MCP servers and database connectors from verified repositories under the awslabs and aws-samples organizations.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided SQL and database metadata, which present a potential attack surface. This is mitigated through a custom SQL builder (safe_query.py) and a linter (dsql_lint) that provide strict validation and sanitization. Ingestion points include user input fields for SQL strings and database results. Sanitization is performed using regex and allowlist validation via the provided Python helper. Capabilities include database mutation and system-level connectivity scripts.
  • [SAFE]: The skill demonstrates high security maturity by implementing proactive defenses against SQL and prompt injection while adhering to official AWS connectivity and credential management patterns.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 10, 2026, 04:14 AM