Developed and integrated a stolen CPU cycles metric into the sys_stats data source for the google/perfetto repository, enabling enhanced visibility into host-steal time for virtualization workloads. The work involved updating Protocol Buffers definitions and parsing logic to capture and report the new steal_ns metric while maintaining compatibility with existing data models and reporting pipelines. Leveraging C++ and Python, the implementation focused on performance analysis and system programming, resulting in improved observability and faster root-cause analysis of CPU contention. Delivered as a targeted, single-commit change, the feature supports more accurate capacity planning without disrupting established workflows or code quality standards.