simready-foundation-conform-fet-005-simulate-grasp-physics

Pass

Audited by Gen Agent Trust Hub on Jul 3, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes local Python scripts (author_grasp_line.py, render_grasp_preview.py) and environment-provided tools (validate-simready-profile) via the uv runner. These operations are required for the skill's primary function of USD asset authoring and validation and are properly scoped to the skill's context.
  • [PROMPT_INJECTION]: The skill processes untrusted USD assets, presenting an indirect prompt injection surface. However, it incorporates robust instruction-level mitigations, requiring vision-capable agents to verify changes against visual evidence and prohibiting automated repairs based on metadata alone.
  • Ingestion points: USD files and source mesh data.
  • Boundary markers: Strict instructions to block operations if visual evidence cannot be inspected directly by the vision-capable agent.
  • Capability inventory: File read/write and script execution via the USD API and uv runner.
  • Sanitization: Use of official industry-standard libraries (pxr) and visual verification steps as part of the core workflow.
  • [SAFE]: No security issues such as data exfiltration, obfuscation, or unauthorized access were detected. The skill originates from a known vendor (nvidia) and follows secure authoring practices for SimReady asset management.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 3, 2026, 11:05 AM
Security Audit — agent-trust-hub — simready-foundation-conform-fet-005-simulate-grasp-physics