omniverse-ui-apis

Pass

Audited by Gen Agent Trust Hub on Sep 1, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface by instructing the agent to ingest and reason about external hierarchical data from USD (Universal Scene Description) files and USDA strings.
  • Ingestion points: Data enters the agent's context through renderer.add_usd() and renderer.add_usd_layer() methods as described in references/inputs-windows-viewport.md and references/recipes.md.
  • Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions when the agent is processing prim names, metadata, or labels derived from untrusted USD content.
  • Capability inventory: The agent has the ability to construct UI components, simulate mouse and keyboard interactions via omni.ui.testing, and perform renderer operations.
  • Sanitization: While the skill advises normalizing path strings for file drops, it lacks specific guidance on sanitizing or escaping content from USD layers before the agent uses that content to populate UI elements or make decisions.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 1, 2026, 02:50 PM
Security Audit — agent-trust-hub — omniverse-ui-apis