What problem does it solve?
Scaffolding a new agentic simulation environment in the Isaac for Healthcare (i4h) workflows repository requires coordinating five separate files (env YAML, assets, task, env class, runtime) with strict conventions, and mistakes only surface at Isaac Sim build time. This Skill guides an agent through forking the closest existing env and producing a runnable first version without inventing new architecture.
Core Features & Use Cases
- Contract-driven scaffolding: Produces exactly the five required env surfaces (YAML config, assets, task, environment class, runtime) under
workflows/agentic/ with repo-root-relative paths.
- Pattern forking and hybrid rules: Selects the closest source env (e.g.,
scissor_pick_and_place, locomanip_tray_pick_and_place), resolves robot owner, policy stack (GR00T N1.5/N1.6, openpi), cameras, and success rules, including G1 hybrid layout and footprint-clearance rules.
- Mandatory validation gate: Runs static checks (
py_compile, --list-envs, dry-runs) plus a real Isaac Sim bridge build with geometry probes and viewport capture inspection before reporting the env ready.
- Use Case: Ask the agent to create a G1 surgical tool sorting env from
scissor_pick_and_place; it loads the recipe, writes the five files with seed layout values, then validates the scene in the bridge.
Quick Start
Ask your agent to create a new i4h environment for surgical tool sorting using the G1 robot based on scissor_pick_and_place.
