iterative-retrieval

Progressive context retrieval is a pattern for gathering and evaluating candidate files based on broad queries, then refining search criteria using discovered patterns and keywords. This iterates up to three cycles to provide sufficient relevant context for subagent tasks.

You refine context retrieval by iteratively evaluating candidate files from initial broad queries, then updating search patterns, keywords, and exclusions based on discovered gaps. This dynamic adjustment prevents overwhelming subagents by pinpointing only the most relevant code snippets.

Use dynamic query adjustment to discover relevant files. The skill evaluates initial broad search results, learns from identified patterns and keywords, and refines the search criteria across up to three iterative cycles to align with the actual codebase terminology.

Yes, iterative retrieval is designed for codebase navigation in complex software engineering tasks. It progressively gathers and evaluates candidate files, dynamically adjusting search patterns and keywords to ensure subagents receive the precise context needed.

The three-cycle maximum limits the depth of iterative search refinement for subagent context retrieval. This constraint bounds the evaluation process, meaning that if sufficient context is not found within three progressive cycles, the search will stop and return the current best results.