This server has a single tool with critical definition gaps. The tool 'execute_skill_action' has a generic, uninformative description (58 chars) that does not explain what the tool actually does, when to use it, or what it returns. No input schema is visible in the source code, the tool appears to be declared only as a string in a JSON status message, not as a proper MCP tool registration with parameters, types, or constraints. The client.py file shows a method 'map_computation_graph_to_dataflow_rdu' with two parameters (model_graph_onnx_uri, memory_bandwidth_tb_s), but these are never connected to the MCP tool definition. The mcp_server.py does not register any tools with schemas, descriptions, or parameter metadata. This is a stub implementation masquerading as an MCP server.

• Define the tool with a proper MCP tool registration including name, description, and input schema. Use a more specific name like 'map_tensor_core_to_dataflow' or 'compile_model_for_rdu'.
• Write a description 50 - 200 chars that answers: What does this tool do? When should the LLM call it? What does it return? Example: 'Map an ONNX model graph to a reconfigurable dataflow tensor core (RDU) and return compilation ID, throughput, and energy efficiency metrics.'
• Add input parameters with types and descriptions: 'model_graph_onnx_uri' (string, S3 URI to ONNX model file), 'memory_bandwidth_tb_s' (number, on-chip SRAM bandwidth in TB/s, range 1 - 100).
• Document the output schema with all returned fields: rdu_compilation_id (string), spatial_tensor_tile_utilization_pct (number), batch_size_1_inference_tokens_per_sec (number), energy_efficiency_tflops_per_watt (number).
• Add error handling: if the ONNX URI is invalid, return 'Invalid model URI. Expected S3 path like s3://bucket/model.onnx'. If bandwidth is out of range, return 'Memory bandwidth must be 1 - 100 TB/s, got X'.