sql-database-assistant
Pass
Audited by Gen Agent Trust Hub on Nov 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external SQL and DDL files, which presents a surface for indirect prompt injection if those files contain adversarial instructions.
- Ingestion points: The Python scripts
scripts/query_optimizer.py,scripts/schema_explorer.py, andscripts/migration_generator.pyread content from files provided via command-line arguments. - Boundary markers: The skill instructions do not provide delimiters or specific warnings to the agent to ignore instructions that may be embedded within SQL comments or string literals in the target files.
- Capability inventory: The agent can execute the included Python scripts and read their output. The scripts themselves are restricted to static text analysis and do not perform network operations or persistent filesystem changes.
- Sanitization: The scripts do not include logic to detect or sanitize natural language instructions that might be hidden within the SQL code to target the LLM's behavior.
Audit Metadata