What is the subagent context problem?

It occurs when secondary AI agents begin a task without knowing which specific files or patterns are necessary, often resulting in context overflow or missing critical information.

How many iterations does this pattern perform?

The pattern is optimized to run for a maximum of 3 cycles, which is typically the 'sweet spot' for finding high-relevance files without wasting processing time.

Can I use this for any programming language?

Yes, the iterative retrieval pattern is language-agnostic and relies on file patterns and keyword discovery rather than specific syntax rules.

How does the evaluation phase work?

It assigns a relevance score from 0 to 1.0 to each retrieved file; files scoring 0.7 or higher are kept as primary context, while lower scores help refine the next search.

Iterative Context Retrieval

Name: Iterative Context Retrieval
Author: arabicapp

byarabicapp

•

생산성 및 워크플로

Optimizes AI subagent performance by progressively refining code context through an intelligent four-stage retrieval loop.

This skill implements a sophisticated iterative retrieval pattern designed to overcome the context window limitations and 'blind start' problems common in multi-agent workflows. By utilizing a 4-phase cycle—Dispatch, Evaluate, Refine, and Loop—it ensures that subagents receive the most relevant code and documentation without overwhelming their context window or guessing required files. It is particularly effective for navigating large, complex codebases where specific domain terminology and architectural patterns must be discovered incrementally to solve a task accurately.

주요 기능

01Automated relevance scoring (0.0 to 1.0) for candidate files

021 GitHub stars

03Context gap identification to drive targeted follow-up queries

04Four-stage retrieval pipeline: Dispatch, Evaluate, Refine, and Loop

05Dynamic keyword and pattern discovery based on initial search results

06Smart exclusion logic to filter out irrelevant paths and noise

사용 사례

01Navigating large legacy codebases with non-standard naming conventions or complex dependencies

02Debugging cross-service issues where the root cause is hidden across multiple architectural layers

03Implementing new features that must adhere to specific, undocumented internal coding patterns

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add arabicapp/everything-claude-code iterative-retrieval

For use in Claude.ai and ChatGPT

Download Skill