iterative-retrieval

To retrieve relevant codebase context for multi-agent systems, you can use an iterative refinement loop that progressively narrows down files through dispatch, evaluate, and refine phases. This dynamically updates search keywords based on previous results to identify high-relevance files.

The best way to optimize RAG context retrieval for complex software engineering tasks is applying a multi-cycle refinement loop. This mechanism evaluates and refines queries iteratively, ensuring token-efficient retrieval and high-relevance file identification without exceeding single prompt limits.

You prevent subagent token exhaustion during large codebase exploration by implementing iterative retrieval. This approach dynamically analyzes codebase terminology and patterns across multiple cycles, preventing failures caused by missing context or excessive token usage in a single prompt.

Yes, you can use iterative retrieval to fix a complex bug when relevant files are unknown. The skill discovers the correct files by analyzing codebase terminology and patterns, intelligently updating search keywords through a progressive context refinement loop.

You should avoid using a single prompt for codebase exploration when the required context is unknown or too large for a single prompt. In these multi-agent workflow scenarios, an iterative retrieval loop is necessary to dynamically evaluate and gather the required context.