user-role
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill documents standard administrative commands for the zilliz CLI, which is the official tool of the skill's author (zilliztech). The skill incorporates safety steps such as instructing the agent to confirm with the user before dropping users or roles.
- [INDIRECT_PROMPT_INJECTION]: The skill instructions define a surface for potential command injection because user-provided data is interpolated into shell command templates without explicit sanitization guidance. (1) Ingestion points: User-provided values for CLI parameters such as username and password in SKILL.md. (2) Boundary markers: The skill includes guidance to confirm with the user before executing drop commands, though it lacks explicit input delimiters. (3) Capability inventory: Generation and execution of shell commands using the zilliz CLI tool as described in SKILL.md. (4) Sanitization: No explicit instructions are provided for the agent to sanitize or escape input strings before shell execution.
Audit Metadata