write-query

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • Indirect Prompt Injection Surface: The skill reads user descriptions and external warehouse schemas to generate and potentially execute SQL queries.
  • Ingestion points: Processes natural language descriptions provided via arguments and retrieves database structure metadata during schema discovery.
  • Boundary markers: Lacks explicit guidelines instructing the agent to separate schema data or user input strings from core behavioral constraints.
  • Capability inventory: Possesses the capability to inspect data structures and offers to execute generated SQL scripts if a connection is established.
  • Sanitization: The workflow does not explicitly describe validation steps or restriction mechanisms to safeguard against the execution of destructive commands if an injection occurs.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 11:26 PM
Security Audit — agent-trust-hub — write-query