adithya-s-k/rl_envs_101
5 skills94 installs820 starsGitHub
Install
npx skills add https://github.com/adithya-s-k/rl_envs_101Skills in this repo
1Generate Verifiers EnvThis skill scaffolds a Verifiers variant of a reinforcement-learning environment, PrimeIntellect's in-process Python library with no HTTP server or Docker. It wraps a domain module as a toolkit plus standalone tool functions, sets up a vf.ToolEnv rollout and composable vf.Rubric graders, and wires a TRL GRPOTrainer path. A developer uses it for fast local RL-env iteration and training with plain Python tools.20installs2Generate Nemo Gym EnvThis skill scaffolds a NeMo Gym variant of a reinforcement-learning environment, NVIDIA's Ray-based RL gym layer for LLM agents. It wraps a shared domain module into a FastAPI resources server that exposes one POST endpoint per tool plus cookie-based sessions and a post-episode /verify reward grader. A developer uses it to port an env to NeMo Gym or build a NeMo resources server for RL training.19installs3Generate Openenv EnvThis skill scaffolds an OpenEnv variant of a reinforcement-learning environment, Meta's HTTP server that exposes tools via the Model Context Protocol. It wraps a shared domain module into a FastAPI server with FastMCP-decorated tools discovered through list_tools, an optional Gradio UI, and sandbox-backed sessions. A developer uses it to port an env to OpenEnv, add MCP tools, or deploy it as a Docker container or HF Space.19installs4Generate Ors EnvThis skill scaffolds an Open Reward Standard (ORS) variant of a reinforcement-learning environment using the official openreward package. ORS is an HTTP REST plus Server-Sent Events protocol where reward arrives inline with every tool output, unlike post-episode grading. A developer uses it to wrap an env in ORS, add per-call rewards, or deploy to OpenReward.ai or HF Spaces.18installs5Rl Env From Descriptionrl-env-from-description converts a plain-English description of a reinforcement-learning training environment into runnable code across OpenEnv, OpenReward (ORS), Verifiers, and NeMo Gym. A developer uses it to scaffold a new RL env, port an existing env between these frameworks, or design its tools, rewards, and state. It drives a clarifying interview, extracts shared domain logic, then writes per-framework implementations plus rollout smoke tests.18installs