What is the Iterative Retrieval pattern?

It is a 4-stage cycle that progressively refines the search for code context by evaluating initial results and using discovered terms to narrow down the most relevant files.

How many retrieval cycles does it perform?

It typically performs up to 3 cycles, stopping early if high-relevance context (score >= 0.7) is found, ensuring a balance between thoroughness and performance.

When should I use this skill?

Use it when performing complex tasks in large repositories where the necessary files aren't immediately obvious or when agents report missing context.

How does this help with multi-agent workflows?

It prevents 'context blindness' in sub-agents by ensuring they receive the specific files they need without hitting token limits or processing irrelevant data.

Iterative Context Retrieval

Name: Iterative Context Retrieval
Author: oabdelmaksoud

byoabdelmaksoud

0•

개발자 도구

Optimizes codebase context retrieval through a multi-stage evaluation loop to prevent context overflow and missing information in multi-agent workflows.

Iterative Context Retrieval is a sophisticated pattern designed to solve the 'context problem' in AI agent orchestration. When sub-agents are spawned, they often lack sufficient information about the codebase or include too much irrelevant data, leading to failures or high token costs. This skill implements a four-stage lifecycle—Dispatch, Evaluate, Refine, and Loop—to progressively narrow down the most relevant files and code snippets. By learning codebase terminology on the fly and scoring file relevance, it ensures that Claude operates with high-precision context, making it ideal for complex bug fixes, feature implementations, and large-scale code exploration.

주요 기능

010 GitHub stars

02Automatic exclusion of irrelevant paths to minimize token waste

03Dynamic terminology discovery to bridge keyword gaps

04Configurable cycle limits to balance retrieval precision and speed

05Relevance scoring system (0.0 to 1.0) for context filtering

06Four-stage retrieval loop: Dispatch, Evaluate, Refine, and Loop

사용 사례

01Large-scale feature implementation requiring discovery of existing patterns

02Deep bug hunting where the root cause location is unknown across a large repo

03Optimizing token usage in complex multi-agent orchestration tasks

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add oabdelmaksoud/superclaw iterative-retrieval

For use in Claude.ai and ChatGPT

주요 기능

010 GitHub stars

02Automatic exclusion of irrelevant paths to minimize token waste

03Dynamic terminology discovery to bridge keyword gaps

04Configurable cycle limits to balance retrieval precision and speed

05Relevance scoring system (0.0 to 1.0) for context filtering

06Four-stage retrieval loop: Dispatch, Evaluate, Refine, and Loop

사용 사례

01Large-scale feature implementation requiring discovery of existing patterns

02Deep bug hunting where the root cause location is unknown across a large repo

03Optimizing token usage in complex multi-agent orchestration tasks