Skip to content
Back to skills

Deployment And Robots

ASecurity

"Export ProtoMotions trackers, validate deployment input contracts,

  • 247 stars
  • 0 votes
  • 0 copies
  • 3 views
  • Added September 8, 2026
devopsgoangular

Security analysis

A100/100

Pro scans all 7 files and shows the line behind each finding

Scanned September 8, 2026

npx -y skills add VectorSpaceLab/AREX-Skill --skill deployment-and-robots --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Deployment And Robots?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Deployment And Robots
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/vectorspacelab-deployment-and-robots/badge)](https://www.skillsdirectory.com/skills/vectorspacelab-deployment-and-robots)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: deployment-and-robots
description: "Export ProtoMotions trackers, validate deployment input contracts,
  test MuJoCo deployment, and add or troubleshoot robot assets."
disable-model-invocation: true
metadata:
  disco-role: operating
license: Apache 2.0
---

# ProtoMotions deployment and robots

Use this sub-skill for G1 deployment tracker workflows, ONNX export, standalone MuJoCo validation, tracker input semantics, real-robot integration cautions, custom robot configs, MJCF/USD conversion, and cross-simulator robot issues.

## Read first

- `references/deployment-contract.md`: ONNX tracker inputs, frame conventions, MuJoCo validation, and real-robot safety.
- `references/onnx-export.md`: export artifact roles and checkpoint/config requirements.
- `references/custom-robots-and-assets.md`: adding robots, asset requirements, MJCF/USD, and factory registration.
- `references/troubleshooting.md`: deployment, input-frame, robot asset, and simulator migration failures.
- `scripts/inspect_tracker_yaml.py`: summarize a deployment YAML sidecar.
- `scripts/tracker_alignment_smoke.py`: pure NumPy smoke for the heading/reference alignment contract.

## Deployment action pattern

1. Verify the checkpoint directory includes `resolved_configs_inference.pt` or an accepted fallback config.
2. Export the policy to `unified_pipeline.onnx` plus `unified_pipeline.yaml` using a CPU-capable environment with PyTorch and ProtoMotions; no simulator should be needed for export.
3. Inspect YAML input/output metadata before wiring a deployment framework.
4. Validate in standalone MuJoCo before real hardware.
5. Preserve heading alignment, root-local angular velocity frame, action post-processing, acceleration clamp, EMA action filter, and PD target semantics.
6. Require explicit human approval and safety procedures for real robot execution.

## Custom robot action pattern

1. Provide MJCF as the source-of-truth robot asset.
2. Define robot config: body-name mappings, trackable body subset, asset config, default root height, PD controls, and per-simulator params.
3. Register the robot name in the factory.
4. Run factory/config smokes before simulator instantiation.
5. Validate in a random-pose or simple simulation check before training.
6. For IsaacLab, handle MJCF-to-USD conversion and D6 workaround behavior carefully.

## Non-negotiable safety

Real-robot deployment can cause unsafe physical motion. Never skip MuJoCo validation, PD/action post-processing checks, emergency-stop verification, or human authorization when moving from simulation to hardware.

Files in this skill

  • SKILL.md2.5 KB
  • references/custom-robots-and-assets.md2.3 KB
  • references/deployment-contract.md3.1 KB
  • references/onnx-export.md2.2 KB
  • references/troubleshooting.md2.4 KB
  • scripts/inspect_tracker_yaml.py1.4 KB
  • scripts/tracker_alignment_smoke.py2.3 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.