Can it learn project-specific terminology?

Yes, the refinement phase extracts keywords and naming conventions found in the first batch of results to improve the accuracy of subsequent searches.

How does the relevance scoring work?

Files are evaluated on a 0-1 scale where 0.8-1.0 identifies direct implementation, 0.5-0.7 identifies related patterns, and lower scores represent tangential or irrelevant code.

What is the 'context problem' this skill solves?

It addresses the issue where AI subagents either have too little information to perform a task or too much irrelevant data that exceeds their context window limits.

How many cycles does the retrieval process run?

The pattern follows a loop that runs for a maximum of 3 cycles, stopping early if sufficient high-relevance context (relevance score >= 0.7) is gathered.

Iterative Context Retrieval

Name: Iterative Context Retrieval
Author: Nixdorfer

byNixdorfer

•

Productivity & Workflow

Refines codebase context progressively to provide subagents with precise, high-relevance information for complex coding tasks.

Iterative Context Retrieval implements a specialized 4-phase loop—Dispatch, Evaluate, Refine, and Loop—designed to solve the common 'context problem' in multi-agent workflows. By progressively narrowing down relevant files based on discovered terminology and scoring relevance on a 0-1 scale, it ensures Claude subagents receive the exact information they need without hitting token limits. This skill is essential for large-scale projects where initial searches often return too much noise or miss critical logic due to unfamiliar internal naming conventions.

Key Features

01Gap identification to target missing context in subsequent search cycles

024-phase iterative retrieval loop for progressive context discovery

03Dynamic keyword and pattern extraction from initial search results

04Smart exclusion of confirmed irrelevant paths to optimize context window

05Automated relevance scoring (0.0 to 1.0) to filter high-value code

061 GitHub stars

Use Cases

01Onboarding subagents to specific modules by providing focused code subsets

02Locating and fixing bugs in massive codebases with unfamiliar architectures

03Implementing features that rely on discovering implicit project-specific patterns

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add nixdorfer/claudecodetool iterative-retrieval

For use in Claude.ai and ChatGPT

Download Skill