do-and-judge

from NeoLabHQ/context-engineering-kit

Execute a task with sub-agent implementation and LLM-as-a-judge verification with automatic retry loop