Loading ToolBench
MCP server for vision-language-action robotic trajectory planning that executes embodied actions using natural language commands and RGB images
This MCP server has critical definition quality gaps. The single tool (execute_skill_action) has a minimal description (99 characters) that lacks essential context for LLM tool selection. The input schema is present but parameters lack adequate detail. The implementation is incomplete, the mcp_server.py file does not actually register the tool with the MCP protocol, it only prints a JSON status message. The client.py and example_usage.py files are example code, not the actual server implementation. There is no demonstrated tool registration, no response schema documentation, and no error handling guidance. The server appears to be a stub or proof-of-concept rather than a functional MCP implementation.
Execute a robotic skill action based on vision and natural language input
Tool description is too brief (99 chars) and lacks context for LLM selection. Does not explain WHEN to use this tool, what prerequisites exist, or what the output contains.
Input parameters lack detailed descriptions. 'scene_rgb_image_url' and 'natural_language_command' have minimal (1-line) descriptions. Parameters need format constraints, examples of expected input, and validation rules.
Output schema is not documented. The tool description does not specify what fields the response contains. Based on client.py example, returns vla_plan_id, command, action_waypoints_7dof_count, gripper_grasp_force_newtons, collision_free_trajectory_verified, control_frequency_hz, trajectory_telemetry_url, but LLMs have no way to discover this from the tool definition.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 34 | 2026-07-28+ | v2 |
No error handling guidance. Tool description does not explain failure modes (e.g., invalid image URL, malformed command, collision detection failure) or how the LLM should respond.
Incomplete MCP server implementation. The mcp_server.py file does not implement the MCP protocol, it only prints a status message. There is no actual tool registration, no request/response handling, no schema definition in JSON-RPC format.
No idempotency information. The tool modifies robotic state (generates trajectory plans), but does not declare whether repeated calls with identical inputs produce the same result or new artifacts. Agents need to know if retry is safe.
Parameter format constraints missing. 'scene_rgb_image_url' should specify: must be valid HTTPS URL, supported image formats (PNG, JPG, WebP), max file size. 'natural_language_command' should specify: max length, any syntax rules or command vocabulary limits.