iterative-retrieval

To help a subagent find relevant files without initial codebase context, use progressive context retrieval. This pattern iteratively dispatches broad queries, evaluates results, and refines search criteria to discover relevant files and project terminology.

Iterative context retrieval for multi-agent workflows is a pattern that progressively refines context gathering. It loops up to three cycles, evaluating retrieved content and adjusting search criteria to ensure subagents acquire sufficient project knowledge.

To dynamically adjust search patterns for codebase navigation, learn from initial file evaluations to refine keywords and exclusions. This progressive refinement loops up to three times to gather sufficient relevant context for agent decisions.

The best way to build a relevant context set for complex debugging is progressive context refinement. By iteratively searching and evaluating files, the agent progressively narrows down search patterns based on initial findings to pinpoint relevant code.

Yes, a limitation of iterative context retrieval is the maximum cycle count. The pattern loops up to a maximum of three cycles to gather sufficient context, meaning highly dispersed or deeply nested codebase patterns may require additional manual searches.