This service agreement concerns the open specification "ONNX" and its associated tools, covering improvements to the reliability of the library’s operation, efficiency and speed. ONNX is a vendor-neutral, open specification allowing machine-learning models to be moved between training frameworks, toolchains, runtimes, and hardware platforms. Beyond the specification itself, ONNX depends on shared maintenance work (security hardening, release integrity, CI infrastructure, ecosystem compatibility, and governance) that every vendor using ONNX relies on. The primary outcomes of the contracted work will be publicly verifiable security and release improvements in the core ONNX project, including obtaining certification for best practices in the field; targeted maintenance of high-impact ecosystem repositories; strengthening the supply-chain integrity. Realising these outcomes entails reviewing the current codebase of tools, CI/CD setups, and contribution practices, and identifying any shortcomings as per the certification requirements and checks, before researching and implementing solutions to bring the above aspects to compliance, in addition to the governance work required to sustain those results and to foster future development of ONNX, based on an analysis of the challenges and opportunities.