
from isaac-for-healthcare
Record teleoperation episodes (keyboard/leader/VR) into HDF5 for dataset creation.
Record episodes from agentic simulation environments via teleoperation into HDF5 files for dataset collection and human demonstration. The skill wraps the i4h-workflows teleop runner, manages run directories, surfaces device-specific keybindings from logs, and verifies saved HDF5 outputs.
Use to collect human demonstration datasets for training or evaluation, or to run teleoperation sessions for testing environments. Requires the i4h-workflows base repo, a configured virtualenv, and the appropriate teleop device available.
An NVIDIA Isaac for Healthcare skill for recording teleoperation episodes into HDF5 datasets. Well-structured SKILL.md with clear instructions, troubleshooting, and prerequisites. No bundled scripts — the skill drives existing bash scripts from a repo clone. The git clone without pinning a commit/tag is a minor concern, and the skill correctly advises against faking demos without a human operator. Very niche audience (robotics teleoperation with specific NVIDIA hardware).
Security deduction: git clone without pinned ref (-8), shell variable expansion in commands (-4). Quality is good: clear steps, good troubleshooting section, explicit verification. Architecture follows skill spec with frontmatter, but no scripts/ or references/ directories. Usefulness is very niche: requires specific NVIDIA robotics hardware and Isaac Sim setup, solving a narrow use case for a small audience.