Skills
Navigate
skills/inference-sh/skills/agent-ui/Gen Agent Trust Hub

agent-ui

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill describes setup instructions that download components and packages from external networks.
  • Evidence: The instructions include npx shadcn@latest add https://ui.inference.sh/r/agent.json and recommend installing belt-sh/cli.
  • Context: The infrastructure at ui.inference.sh belongs to the skill vendor (inference-sh), matching normal expected operations.
  • [INDIRECT_PROMPT_INJECTION]: The UI component creates a potential surface for indirect prompt injection by aggregating untrusted data while executing browser or backend tools.
  • Ingestion points: Untrusted inputs enter via file and image uploads through the allowFiles and allowImages properties, and text fields in SKILL.md.
  • Boundary markers: Absent within the client-side instantiation examples.
  • Capability inventory: The agent possesses capabilities to invoke client-side tools like scan_ui and fill_field, as well as proxy backend interactions via /api/inference/proxy.
  • Sanitization: Absent or not highlighted in the provided configuration templates.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 01:38 PM
Security Audit — agent-trust-hub — agent-ui