技能创建器
skill-creator
创建、改进和评估自定义技能,把专业流程转化为可重复使用的 AI 能力。
motion-tools【数据与研究】Inspect, convert, validate, view, and plot MimicKit motion, dataset, viewer, DOF, conversion, and log artifacts.
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
Use this sub-skill when a task is about MimicKit motion artifacts rather than policy training: motion pickle files, motion dataset manifests, GMR or SMPL/AMASS conversion, view_motion visualization, character DoF diagnostics, and training-log plots.
.npz data to the SMPL humanoid MimicKit motion layout.log.txt-style training logs to an image using a headless-safe default.All helpers are self-contained and accept ordinary file paths. When a target MimicKit checkout should be made importable, pass its root explicitly with --repo-root; do not rely on the current working directory being the checkout.
runner-and-backends sub-skill.smp sub-skill; this sub-skill only covers generic motion/data conversion and visualization.The generated guidance is backed by source inspection plus lightweight checks: CPU/CUDA torch imports, parser help for the GMR and SMPL converters, source compile checks, and tiny converter fixtures. Simulator-native viewing and DOF-test workflows still require an external supported simulator backend plus downloaded motions/models/assets before they can be fully validated.