i built a skill that puts 45 sub-agents for every important task... like real experts, their real frameworks, all documented... not some generic "panel of experts" prompt that's just claude acting and hallucinating here's how it works: > i call the skill on whatever i'm working on > it figures out which experts are relevant to THIS specific task > spins up a sub-agent for each one > they evaluate my work through their own lens these experts don't agree with each other, i carefully picked…