vision-expert

Pass

Audited by Gen Agent Trust Hub on Feb 18, 2026

Risk Level: SAFENO_CODE
Full Analysis
  • [Prompt Injection] (SAFE): No malicious instructions, bypass attempts, or role-play injections were detected. The instructions are standard for a vision-to-code task.
  • [Data Exposure & Exfiltration] (SAFE): No access to sensitive file paths (~/.ssh, .env) or unauthorized network operations were identified.
  • [Remote Code Execution] (SAFE): No external scripts, remote downloads, or package installations are present in the skill definition.
  • [Indirect Prompt Injection] (LOW): The skill processes untrusted visual data (screenshots/diagrams). While visual inputs can carry instructions, the skill's scope is restricted to 'zai-vision' tools and it lacks permissions for shell execution or file-system writing, which limits the potential impact of an injection.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 18, 2026, 07:36 PM