You need a full software team to ship a product, for agents it is not different.
Clear roles, handoffs, automated tests, CI/CD, reviews will have to run in loops. A single agent is weak. Agents checking other agents will massively improve multiagent development.
8/ Set up long-lived sessions with roles: planner, implementer, adversarial reviewer, black-box tester, issue triager, deep code reviewer. Your job is to wire them together so the implementer is never idle, and monitor and review to make sure everything is working and catch mistakes.