i4h-workflow-finetune

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill clones the i4h-workflows repository from https://github.com/isaac-for-healthcare/i4h-workflows. This is a vendor-owned resource (NVIDIA) and is the official source for the Isaac for Healthcare workflow scripts and configuration.
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands for repository management (git clone), environment setup, and policy training (uv run). These commands are standard for the development workflow and are restricted to the local workspace.
  • [INDIRECT_PROMPT_INJECTION]: The skill has a data ingestion surface through external robotic datasets.
  • Ingestion points: Reads dataset configuration and metadata from meta/info.json within the user-provided DATASET_PATH.
  • Boundary markers: None identified; the skill assumes the provided path follows the LeRobot dataset structure.
  • Capability inventory: File system access (reading metadata) and shell command execution (git, uv).
  • Sanitization: The skill verifies the existence of the expected metadata file (meta/info.json) before proceeding with the workflow.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 07:51 AM
Security Audit — agent-trust-hub — i4h-workflow-finetune