🛡️ OWASP LLM02 • SQL Injection Defense • Insecure Output Handling • AST Auditing
SQL AST Security Sanitizer & Guardrail
Never execute raw SQL generated by LLMs without deterministic AST inspection. Intercept state-mutating commands (DROP, DELETE, UPDATE) and enforce mandatory pagination limits.
LLM SQL Output to Audit
AST Guardrail Rules:
🛡️ Defense-in-Depth Principle
Always pair application-level AST validation with read-only database user roles (`GRANT SELECT ON ALL TABLES IN SCHEMA public TO ai_read_only_agent`).
AST Security Verdict
🔴 Blocked: Dangerous DDL / Multi-Statement Violations Detected:
Sanitized Executable SQL (Safe for Engine):