iterative-retrieval

Iterative context retrieval refines context dynamically by using a 4-phase loop: DISPATCH, EVALUATE, REFINE, and LOOP. This allows subagents to progressively adjust and retrieve relevant context during task execution.

Iterative retrieval for codebase exploration works by dispatching subagents to evaluate initial context, refining their queries based on evaluation results, and looping until the relevant code context is gathered without redundancy.

Dynamic context adjustment is needed when subagents cannot determine the required context until they begin their tasks. This is common in complex codebase exploration where dependencies and relevant files are discovered during execution.

Yes, you can use iterative retrieval to design retrieval pipelines for code exploration. It supports scenarios requiring progressive context refinement, enabling agents to dynamically adjust what codebase context they retrieve.

The best way to reduce context redundancy is using an iterative retrieval loop. By continuously evaluating and refining the context fetched during multi-agent interactions, the workflow discards irrelevant data and enhances task relevance.

No specific dependencies are required to use iterative retrieval for context management. The skill operates independently to provide the progressive context refinement loop for multi-agent workflows.