motrix-rl-skrl owns MotrixLab's SKRL integration: provider configuration, environment wrappers, PPO trainers and export
implementation. Its Python namespace is motrix_rl_skrl; typed configuration is motrix_rl_skrl.config.SkrlCfg.
The method is skrl.ppo, with Torch or JAX training backends. Installing motrix-rl-skrl includes SKRL and Torch;
motrix-rl-skrl[jax] adds the Linux-only JAX runtime. Workspace installation is managed by
sh install.sh --skrl-torch or sh install.sh --skrl-jax, which also selects the workspace GPU wheel profile.
The plugin depends on motrix-rl control-plane services and contracts, SKRL, and Torch. Only the additional JAX backend
libraries are optional within the plugin. This plugin is a required workspace root dependency for the built-in PPO demos;
the core package never depends on it. Discovery uses the motrix_rl.frameworks entry-point group:
skrl = "motrix_rl_skrl.plugin:register". Registration declares provider capabilities and installs the algorithm
schema before Hydra task composition; it does not create a parallel task registry.