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