When does the retrieval process stop?

The process stops when the agent finds at least three high-relevance files (scored 0.7 or higher) or completes a maximum of three refinement cycles.

Can this skill help with large codebases?

Yes, it is specifically designed for large repositories where sending all files to the AI is impossible due to context window limitations.

How does iterative retrieval improve AI performance?

By searching in cycles, the AI can find project-specific terms (like 'throttle' vs 'rate-limit') in the first pass and use those terms to find the exact code needed in the second pass.

What is the sub-agent context problem?

It occurs when autonomous agents are started without knowing which files are relevant, leading them to either guess, miss critical info, or exceed context limits by trying to read everything.

Iterative Context Retrieval

Name: Iterative Context Retrieval
Author: Fabio29T

byFabio29T

0•

개발자 도구

Optimizes codebase search through a multi-stage refinement process to provide sub-agents with high-relevance context.

The Iterative Retrieval skill implements a structured four-stage pattern—Dispatch, Evaluate, Refine, and Loop—to solve the common 'context problem' in multi-agent workflows. When sub-agents are tasked with complex coding challenges, they often lack the initial knowledge of which files are relevant or what specific terminology the codebase uses. This skill allows Claude to progressively narrow down its search, scoring file relevance on a 0-1 scale and learning project-specific patterns dynamically. By iterating up to three times, it ensures the agent receives precise, high-value context while avoiding the pitfalls of exceeding token limits or operating with insufficient data.

주요 기능

01Relevance Scoring: Evaluates and ranks file content (0-1) to ensure only critical context is used.

02Dynamic Terminology Discovery: Learns project-specific naming conventions during initial broad queries.

034-Stage Retrieval Cycle: Structured approach to finding code through progressive refinement.

04Context Window Management: Prevents token overflow by filtering out low-relevance files and patterns.

05Gap Identification: Explicitly identifies missing information to drive more accurate subsequent searches.

060 GitHub stars

사용 사례

01Onboarding sub-agents to large codebases where architectural patterns aren't immediately obvious.

02Debugging complex issues where the root cause is hidden across multiple unfamiliar files.

03Implementing new features that require gathering specific middleware, routing, and utility patterns first.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add fabio29t/everything-claude iterative-retrieval

For use in Claude.ai and ChatGPT

주요 기능

01Relevance Scoring: Evaluates and ranks file content (0-1) to ensure only critical context is used.

02Dynamic Terminology Discovery: Learns project-specific naming conventions during initial broad queries.

034-Stage Retrieval Cycle: Structured approach to finding code through progressive refinement.

04Context Window Management: Prevents token overflow by filtering out low-relevance files and patterns.

05Gap Identification: Explicitly identifies missing information to drive more accurate subsequent searches.

060 GitHub stars

사용 사례

01Onboarding sub-agents to large codebases where architectural patterns aren't immediately obvious.

02Debugging complex issues where the root cause is hidden across multiple unfamiliar files.

03Implementing new features that require gathering specific middleware, routing, and utility patterns first.