// Execute or plan a real task on Piper, Unitree G1, or Unitree Go2. Use when the user asks the robot fleet to do something in the physical world, including observing, moving, manipulating, inspecting, or multi-robot task execution.
External services and backend capabilities for AbotClaw robots — vision APIs, VLM/LLM endpoints, planners, speech services, grasping backends, and client SDKs. Use when a robot skill needs an external service, when checking what backend tools already exist, when debugging service reachability, or when deciding whether to build locally versus call a service.
Bundle an AbotClaw robot skill and its dependencies into a single executable Python file or portable submission artifact. Use when preparing robot code for execution environments that want one file, when flattening dependency chains, or when packaging a skill for deployment or review.
SpatialMemoryHub-based robot memory querying, writing, and retrieval for AbotClaw. Use when the agent needs to store or retrieve object memory, place memory, keyframe memory, or semantic frame memory; when checking whether a memory service is running; when using robot SDK camera frames to perform memory ingestion or memory-guided retrieval; or when grounding robot tasks in prior visual/spatial experience.
Use a deployed VLAC-style vision-language-action critic service to evaluate task progress, compare current observations against a reference image, and judge task completion from robot camera frames. Use when the agent needs external progress supervision, completion verification, failure detection, or image-based task-state comparison for Piper, Unitree G1, or Unitree Go2.
Robot connection and reachability guidance for the AbotClaw fleet. Use when identifying the correct robot base URL, auth method, health check, safest first connectivity test, or when debugging why Piper, Unitree G1, or Unitree Go2 cannot be reached.
Hardware roles, embodiment boundaries, and task-fit guidance for Piper, Unitree G1, and Unitree Go2. Use when deciding which robot should handle a task, when reasoning about embodiment constraints, or when documenting robot-specific assumptions in a skill.
| name | abotclaw-run-robot-task |
| description | Execute or plan a real task on Piper, Unitree G1, or Unitree Go2. Use when the user asks the robot fleet to do something in the physical world, including observing, moving, manipulating, inspecting, or multi-robot task execution. |
Follow this order. Do not jump straight into code.
Decide whether the request is mainly:
Pick Piper, G1, Go2, or a staged combination.
Before new code, use SDK discovery:
If the interface is not yet known, stop and discover it first.
Look for existing local skills or reusable code paths before building from scratch.
Because there is no simulator safety layer here:
The first successful robot action should usually be a small probe that confirms the API and embodiment assumptions, not a heroic end-to-end attempt.
