Package, quantize, validate, and publish OpenRAL rSkills to Hugging Face Hub.
npx skills add https://github.com/bensonlee5/openral --skill rskill-packager
Please help me install this Agent Skill. Skill: rskill-packager Source: https://github.com/bensonlee5/openral/tree/main/.agents/skills/rskill-packager Command: npx skills add https://github.com/bensonlee5/openral --skill rskill-packager
The rskill-packager Skill solves the problem of creating, quantizing, validating, and publishing OpenRAL rSkills. It ensures that the packaging adheres to schema, license, safety, and reproducibility rules.
Pack and publish an rSkill for 'pi05-pick-cube' from Hugging Face Hub with 'openral rskill new pi05-pick-cube --from-hf <owner>/<repo>'.
High-intent search queries and answers about installing and using this skill.
To package and publish an OpenRAL rSkill, you can scaffold a new package using `openral rskill new`, validate schema and license compliance, and push the result directly to Hugging Face Hub using the integrated publishing tools.
Quantization for OpenRAL rSkills converts model checkpoints into model-safe tensors. You need this process to compress model size and ensure safe, reproducible execution before publishing your rSkill package to the Hub.
Yes, you need an OpenRAL runtime and Hugging Face Hub access to validate rSkills. The validation process checks README quality, eval artifacts, and license posture, requiring these environments to function correctly.
Yes, you can scaffold a new rSkill package from a Hugging Face repository. Use the command `openral rskill new <name> --from-hf <owner>/<repo>` to fetch and populate the initial package structure directly.
Before publishing an OpenRAL rSkill, validation checks include README quality, eval artifacts, license posture, safety, and reproducibility. These gates ensure your package adheres to required schema and deployment rules.
evalstateSubmit TRL training jobs on Hugging Face Jobs and convert models to GGUF.
bensonlee5Automate pen pick-and-place tasks on a real SO-101 follower arm using a VLA policy.
bensonlee5Run pick and place operations on RoboCasa GR-1 tabletop tasks.
NVIDIA-NeMoVerify weight parity between HF and MCore model conversions using Python scripts.
