Claude has now release agent swarm to split up a feature development into multiple tasks, check dependency on other tasks and use sub agents to not fill up the context window.
I believe this workflow improves the quality at least 3X if not 10X
As Dean mentioned, subagents are what you are looking for. I personally created a skill for planning an implementation, which results in a plan that accounts for parallel implementation. Then I have a skill for the parallel implementation, which splits all work across ~6 agents. I could do more, but I think this is more than necessary.