How does it help with token limits?

Instead of sending every possible file to an agent, it evaluates each file for relevance and only includes the highest-scoring content, preventing 'context overflow' and reducing costs.

How many cycles does the retrieval loop perform?

The pattern typically recommends a maximum of 3 cycles to find sufficient context, ensuring a balance between thoroughness and execution speed.

When should I use this skill?

Use it when building multi-agent workflows, exploring large repositories, or when you notice agents failing due to missing information or 'guessing' project patterns.

What is the Iterative Retrieval skill?

It is a specialized capability for Claude Code that helps sub-agents find the exact codebase context they need by using a repeating search-and-refine process rather than a single search.

Does this work with any programming language?

Yes, the iterative retrieval pattern is language-agnostic as it relies on file patterns, keywords, and relevance scoring applicable to any codebase.

Iterative Retrieval

Name: Iterative Retrieval
Author: saar120

bysaar120

0•

生产力与工作流

Optimizes sub-agent performance by incrementally discovering and refining codebase context through a multi-step retrieval loop.

The Iterative Retrieval skill addresses the 'context problem' in multi-agent workflows where sub-agents often lack the specific codebase knowledge needed to perform a task accurately. By implementing a structured 4-step loop—Dispatch, Evaluate, Refine, and Loop—this skill enables Claude to start with broad searches and progressively narrow down to the most relevant files and patterns. This approach prevents token limit issues caused by sending too much data while ensuring agents have the critical implementation details, project terminology, and structural patterns required for high-quality code generation and debugging.

主要功能

01Token usage optimization through selective context inclusion

02Intelligent gap identification to guide subsequent search cycles

034-stage context refinement loop (Dispatch, Evaluate, Refine, Loop)

04Codebase terminology discovery to adapt searches to project-specific naming

05Automated relevance scoring for retrieved code files (0.0 to 1.0)

060 GitHub stars

使用场景

01Fixing complex bugs that require context across multiple interconnected modules

02Navigating large, unfamiliar repositories where initial keyword searches are insufficient

03Implementing new features that must strictly adhere to existing architectural patterns

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add saar120/skillgate-registry iterative-retrieval

For use in Claude.ai and ChatGPT

主要功能

01Token usage optimization through selective context inclusion

02Intelligent gap identification to guide subsequent search cycles

034-stage context refinement loop (Dispatch, Evaluate, Refine, Loop)

04Codebase terminology discovery to adapt searches to project-specific naming

05Automated relevance scoring for retrieved code files (0.0 to 1.0)

060 GitHub stars

使用场景

01Fixing complex bugs that require context across multiple interconnected modules

02Navigating large, unfamiliar repositories where initial keyword searches are insufficient

03Implementing new features that must strictly adhere to existing architectural patterns

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add saar120/skillgate-registry iterative-retrieval

For use in Claude.ai and ChatGPT