Feature Request: Add support for agent to agent communication

Now that subagents are supported out of the box (thank you by the way! was doing all of that manually before), it is great to have them work on things in order. Eg: A developer sub agent writes the code, a testing subagent tests the code and writes automation.

However, there are times where I want more back and forth between my agents, similar to how real engineers operate.
For example, the developer writes the code, the tester tests the code against acceptance criteria and determines that the code is not working as intended, the tester then sends a message to the developer letting them know the tests failed, and here are the ones that failed with specific details.

Another recent example use case: I decided to put claude code's limits to the test and throw an entire new epic at the system. This involved setting up a new repo from scratch, based on a template, and then implementing several api endpoints and workflows. I manually had an architecture subagent design out how I wanted everything to be done, and then asked claude code to spin up multiple developer subagents to tackle each part of the epic that could be done independently.
Things worked well for a while, until each one started trying to build the code, realized there were tons of files added that it was not aware of and that had nothing to do with that agent's work, so the next requests from every one of the agents was some form of rm AnotherAgentsFiles.cs to "fix" the bad build. If they had a form of communication, they could have known that everyone of them was working on a different piece of the puzzle, and to not destroy the other's work.

These are simple examples, but there is much more power in being able to throw a fully functional team that can communicate at some of our tasks! Please add this to claude code