provision-hyperscale
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- Indirect Prompt Injection: The skill provides commands that interpolate user-supplied parameters into shell commands, creating a potential surface for injection if inputs are not validated. \n
- Ingestion points: User-provided database names, server names, and resource group names as described in
SKILL.md. \n - Boundary markers: No specific delimiters are used to wrap user-supplied data in the provided command templates. \n
- Capability inventory: The skill uses
az sqlandsqlcmdto manage cloud database resources and configurations (found inSKILL.md). \n - Sanitization: No input sanitization logic is included in the instruction set.
Audit Metadata