How many cycles does the retrieval pattern run?

The pattern is designed to run for a maximum of 3 cycles, which provides a balance between finding the right context and maintaining performance.

Can it adapt to different coding styles?

Yes, the refinement phase specifically looks for project-specific terminology (e.g., 'throttle' instead of 'rate-limit') to improve search accuracy in later cycles.

How does the scoring system work?

Files are scored from 0 to 1; files with a score of 0.7 or higher are prioritized, while those below 0.2 are excluded from the context.

What problem does Iterative Retrieval solve?

It solves the 'context problem' where AI agents either receive too much irrelevant information or too little specific data to solve a task effectively.

Iterative Retrieval Pattern

Name: Iterative Retrieval Pattern
Author: kimliss

bykimliss

0•

Herramientas para Desarrolladores

Optimizes codebase context retrieval by progressively refining search criteria through an intelligent multi-phase evaluation loop.

The Iterative Retrieval skill addresses the 'context window' challenge in multi-agent AI workflows where subagents often lack sufficient codebase knowledge to perform complex tasks. By implementing a sophisticated four-phase cycle—Dispatch, Evaluate, Refine, and Loop—it enables Claude to autonomously discover relevant files, learn project-specific terminology, and identify critical gaps. This pattern ensures that agents operate with the most pertinent information while avoiding context overflow, making it an essential utility for navigating large-scale repositories during bug fixes or feature implementations.

Características Principales

01Four-phase autonomous retrieval loop (Dispatch, Evaluate, Refine, Loop)

020 GitHub stars

03Intelligent relevance scoring (0-1 scale) to eliminate codebase noise

04Built-in cycle limits to balance retrieval depth with execution speed

05Explicit gap identification to guide secondary search queries

06Dynamic terminology discovery to adapt to project-specific naming conventions

Casos de Uso

01Reducing token consumption by filtering out irrelevant files during complex agent tasks

02Navigating large legacy codebases where file relationships are not immediately obvious

03Pinpointing root causes for bugs that span across multiple interconnected modules

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add kimliss/claude-code-inhand iterative-retrieval

For use in Claude.ai and ChatGPT

Download Skill