AWS releases coding-agent skill for SageMaker inference benchmarking

AWS has introduced the aws-ai-ml skill through its Agent Toolkit, enabling compatible coding agents to generate SageMaker inference benchmarking and configuration code. AWS says the skill supports endpoint load tests, instance recommendations and comparisons between performance runs, using executable SageMaker Python SDK v3 code that engineers can inspect and run. The skill works with agents supporting the Model Context Protocol, which connects agents to tools, and is also available in a preconfigured SageMaker Studio environment. Benchmarks drive real traffic to live endpoints and require confirmation. The skill generates deployment configurations rather than deploying models itself.