AGENTS with SKILLS
each on a different $model
save the expensive models and do most work via delegation (clean context) and 💥

my token burn is down 50%+ rarely hit limits now, I downgraded 🤣