To an extent. I've had GPT 5.5 solve problems that Opus 4.7 struggled with, using an identical AGENTS.md/CLAUDE.md and no skills.