npx claudepluginhub plurigrid/asi --plugin asiThis skill uses the workspace's default tool permissions.
**Trit**: 0 (ERGODIC - coordination/infrastructure)
Provides hardware specifications, MJCF simulation models, and deployment configurations for K-Scale Labs' K-Bot humanoid robot. Useful for robotics simulation, control, and integration.
Guides training RL agents with Stable Baselines3 (PPO, SAC, DQN, TD3, DDPG, A2C), custom Gym environments, callbacks for monitoring, vectorized envs for parallel training, and deep RL workflows.
Guides training RL agents with Stable Baselines3 algorithms (PPO, SAC, DQN, TD3, A2C) using Gymnasium environments, custom env creation, callbacks, and optimization.
Share bugs, ideas, or general feedback.
Trit: 0 (ERGODIC - coordination/infrastructure) Color: #5B8DEE (Sky Blue) URI: skill://kscale#5B8DEE
This skill indexes the K-Scale Labs robotics ecosystem - a comprehensive open-source stack for building, training, and deploying humanoid robots. The collection follows GF(3) triadic organization with kos-firmware (+1) as the primary generator and mujoco-scenes (0) as the coordinator.
┌────────────────────────────────────────────────────────────────────┐
│ K-SCALE SKILL ECOSYSTEM │
├────────────────────────────────────────────────────────────────────┤
│ │
│ PLUS (+1) - Generation/Construction │
│ ┌─────────────────────────────────────────────────────────────┐ │
│ │ kos-firmware #79ED91 Robot firmware & gRPC services │ │
│ └─────────────────────────────────────────────────────────────┘ │
│ │
│ ERGODIC (0) - Coordination/Infrastructure │
│ ┌─────────────────────────────────────────────────────────────┐ │
│ │ mujoco-scenes #9FD875 Scene composition for MuJoCo │ │
│ └─────────────────────────────────────────────────────────────┘ │
│ │
│ MINUS (-1) - Analysis/Verification │
│ ┌─────────────────────────────────────────────────────────────┐ │
│ │ ksim-rl #3A2F9E RL training for locomotion │ │
│ │ evla-vla #DBA51D Vision-language-action model │ │
│ │ urdf2mjcf #4615B7 URDF to MJCF conversion │ │
│ │ kbot-humanoid #5B45C2 K-Bot robot specifications │ │
│ │ zeroth-bot #8CC136 3D-printed humanoid platform │ │
│ │ kscale-actuator #B9172E Robstride motor control │ │
│ │ entropy-sim2real #E85B8E Entropy-driven sim2real │ │
│ └─────────────────────────────────────────────────────────────┘ │
│ │
└────────────────────────────────────────────────────────────────────┘
Current State: UNBALANCED (trit sum = -6)
| Trit | Count | Skills |
|---|---|---|
| +1 (PLUS) | 1 | kos-firmware |
| 0 (ERGODIC) | 1 | mujoco-scenes |
| -1 (MINUS) | 7 | ksim-rl, evla-vla, urdf2mjcf, kbot-humanoid, zeroth-bot, kscale-actuator, entropy-sim2real |
The primary balanced triad anchors the ecosystem:
ksim-rl (-1) ⊗ kos-firmware (+1) ⊗ mujoco-scenes (0) = 0 ✓
Pattern: Each MINUS skill can form a balanced triad by reusing the (kos-firmware, mujoco-scenes) pair:
evla-vla (-1) ⊗ kos-firmware (+1) ⊗ mujoco-scenes (0) = 0 ✓
urdf2mjcf (-1) ⊗ kos-firmware (+1) ⊗ mujoco-scenes (0) = 0 ✓
kbot-humanoid (-1) ⊗ kos-firmware (+1) ⊗ mujoco-scenes (0) = 0 ✓
zeroth-bot (-1) ⊗ kos-firmware (+1) ⊗ mujoco-scenes (0) = 0 ✓
kscale-actuator (-1) ⊗ kos-firmware (+1) ⊗ mujoco-scenes (0) = 0 ✓
entropy-sim2real (-1) ⊗ kos-firmware (+1) ⊗ mujoco-scenes (0) = 0 ✓
To achieve independent balanced triads (no skill reuse), add PLUS (+1) skills:
┌─────────────────────┐
│ TRAINING │
│ ┌───────────────┐ │
│ │ ksim-rl │ │
│ │ (PPO, AMP) │ │
│ └───────────────┘ │
│ │ │
│ ▼ │
│ ┌───────────────┐ │
│ │ mujoco-scenes │ │
│ │ (environments)│ │
│ └───────────────┘ │
└─────────┬───────────┘
│
┌──────────────────────┼──────────────────────┐
│ │ │
▼ ▼ ▼
┌─────────────────┐ ┌─────────────────┐ ┌─────────────────┐
│ MODELS │ │ PERCEPTION │ │ TRANSFER │
│ ┌─────────────┐ │ │ ┌─────────────┐ │ │ ┌─────────────┐ │
│ │ urdf2mjcf │ │ │ │ evla-vla │ │ │ │entropy-sim2r│ │
│ │ kbot-humanoi│ │ │ │ (VLA) │ │ │ │ (domain │ │
│ │ zeroth-bot │ │ │ └─────────────┘ │ │ │ random.) │ │
│ └─────────────┘ │ └─────────────────┘ │ └─────────────┘ │
└─────────────────┘ └─────────────────┘
│ │
└──────────────────────┬──────────────────────┘
│
▼
┌─────────────────────┐
│ DEPLOYMENT │
│ ┌───────────────┐ │
│ │ kos-firmware │ │
│ │ (+1) │ │
│ └───────────────┘ │
│ │ │
│ ▼ │
│ ┌───────────────┐ │
│ │kscale-actuator│ │
│ │ (CAN motors) │ │
│ └───────────────┘ │
└─────────────────────┘
# Full K-Scale training pipeline
from ksim import PPOTask
from ksim.robots.kbot import KBotConfig
from mujoco_scenes import SceneBuilder, Terrain
class KBotWalkingTask(PPOTask):
robot = KBotConfig(model_path="kbot.mjcf")
def build_scene(self):
scene = SceneBuilder()
scene.add_terrain(Terrain.FLAT)
scene.add_random_obstacles(count=5)
return scene.to_mjcf()
# Entropy-driven domain randomization
physics_randomizers = [
StaticFrictionRandomizer(scale=0.5),
MassMultiplicationRandomizer(scale=0.2),
]
# Train
task = KBotWalkingTask()
task.run_training(num_envs=4096)
from pykos import KosClient
async def deploy_policy():
async with KosClient("kbot.local:50051") as client:
# Load trained policy
await client.policy.load("walking_v1.onnx")
# Start control loop
await client.policy.start()
# Monitor
while True:
state = await client.actuator.get_actuators_state()
print(f"Position: {state.positions}")
| Skill | Repository |
|---|---|
| ksim-rl | kscalelabs/ksim |
| kos-firmware | kscalelabs/kos |
| evla-vla | kscalelabs/evla |
| urdf2mjcf | kscalelabs/urdf2mjcf |
| kbot-humanoid | kscalelabs/kbot |
| zeroth-bot | kscalelabs/zeroth-bot |
| mujoco-scenes | kscalelabs/mujoco-scenes |
| kscale-actuator | kscalelabs/actuator |
jaxlife-open-ended (+1): Open-ended evolution for behavior discoveryergodicity (0): Ergodic theory foundationswobble-dynamics (0): Perturbation response analysisstability (-1): Dynamical system stability analysis@misc{kscale2024,
title={K-Scale Labs Open Source Robotics Stack},
author={K-Scale Labs},
year={2024},
url={https://github.com/kscalelabs}
}