iterative-retrieval

Progressive context refinement loop for multi-agent workflows solving the subagent context problem. Solves the "context problem" in multi-agent workflows where subagents don't know what context they need until they start working. Subagents are spawned with limited context. They don't know: Standard approaches fail: A 4-phase loop that progressively refines context: Initial broad query to gather candidate files: Assess retrieved content for relevance: Scoring criteria: Update search criteria based on evaluation: Repeat with refined criteria (max 3 cycles): Use in agent prompts: - Four-phase cycle (dispatch, evaluate, refine, loop) that iteratively narrows retrieval to high-relevance files, capping at 3 iterations to balance token usage and context quality - Scores retrieved files on a 0–1 relevance scale and explicitly identifies missing context gaps to drive the next refinement cycle - Learns codebase terminology and patterns during the first cycle, improving subsequent searches without requiring upfront knowledge of project conventions - Designed for spawning subagents on code exploration, bug fixes, and feature implementation tasks where context needs are unpredictable upfront - Spawning subagents that need codebase context they cannot predict upfront - Building multi-agent workflows where context is progressively refined - Encountering "context too large" or "missing...