Zero2Skill

First analyze the task and generate a collect/reset plan, then loop collect + reset until done. Trigger when the user mentions "analyze task", "step breakdown", or "learn a task".

Use when the user mentions "grasp", "pick up", or "pick and place". Uses SAM3 text segmentation + AnyGrasp planning to drive a robot arm for grasping, optionally place at a given (x,y); supports --collect to record HDF5.

Robot-arm self-learning skill that orchestrates analysis, collection, reset, and related skills in order.

Skill for understanding RGB images from three cameras. Trigger when the user mentions "understand images" or "understand the environment".