iterative-retrieval

Automate context gathering for subagents using a 4-phase loop: DISPATCH, EVALUATE, REFINE, and LOOP. This process progressively filters and scores candidate files to prioritize high-relevance content for your agents.

Subagents waste cycles when they start with limited context and don't know what they need until they begin work. Iterative retrieval solves this by progressively refining context across retrieval backends before task execution.

Use a relevance-based file selection loop to assemble high-value context. It scores candidate files and progressively refines the gathered data across a maximum of three cycles for smarter code exploration.

Progressive context retrieval works by dispatching subagents, evaluating the initial context, refining the search queries, and looping until high-value context is assembled, capping at three cycles to prevent infinite loops.

Yes, it applies to spawning subagents that need progressively refined context across codebases and retrieval backends. It filters and scores candidate files to prioritize high-relevance content for practical use cases like code exploration.

The 4-phase loop is capped at a maximum of three cycles to prevent infinite loops. This limits the depth of context refinement but ensures the workflow terminates efficiently while scoring candidate files for relevance.