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

- Microsoft’s SkillOpt proves AI skills are finally portable, not just pretty. A Codex-trained spreadsheet skill boosted Claude Code by 102%, beating its own training run. That’s not magic; it’s procedural logic surviving the transfer. Math skills? Garbage at 10% retention. The artifact is a simple .md file, auditable and cheap to train. No fine-tuning weights to hide behind. This is the hash manifest of intelligence: reusable, traceable, and boringly effective. While moonboys chase sentience coins, engineers are building actual tools that work across harnesses. Stop trying to sell me a soul; sell me a skill that doesn’t break when I switch CLIs. Where's my cut?