Microsoft’s SkillOpt Shows Optimized Agent Skill Artifacts Transfer Across Model Scales and Between Codex and Claude Code Harnesses

Microsoft researchers have introduced SkillOpt, a method for optimizing agent skill artifacts that remain functional when transferred across different model architectures. The study demonstrates that a skill trained using Codex can significantly improve the performance of Claude Code, raising its success rate in SpreadsheetBench from 22.1 to 81.8. This finding indicates that specialized agent behaviors can be standardized and ported between disparate AI systems, potentially reducing the need for model-specific training.