The Iterative Retrieval skill addresses the 'context problem' in multi-agent workflows where agents lack necessary information at startup. Instead of guessing or overloading the context window, this skill implements a systematic four-phase cycle—Dispatch, Evaluate, Refine, and Loop—to discover relevant files. By scoring results, identifying information gaps, and adjusting search parameters based on discovered terminology, it ensures sub-agents operate with high-precision information while staying within token limits.
주요 기능
01Context gap identification
02Token-efficient context management
03Automated keyword and pattern discovery
04Relevance scoring system (0.0 to 1.0 scale)
051 GitHub stars
06Four-phase iterative discovery loop
