What is considered a high-relevance file?

Files with a score of 0.7 to 1.0, which typically directly implement the target functionality or contain the primary logic required for the task.

What is the Iterative Retrieval pattern?

It is a method for AI agents to progressively find relevant code by searching, scoring results for relevance, identifying information gaps, and refining subsequent searches.

How many cycles does the retrieval process typically take?

The pattern recommends a maximum of 3 cycles. This provides a balance between finding the necessary context and maintaining efficiency/speed.

How does this solve the subagent context problem?

Subagents often start with too little information. This skill gives them a framework to explore the codebase and find what they need before attempting to write code.

Iterative Context Retrieval

Name: Iterative Context Retrieval
Author: ysyecust

byysyecust

•

Developer Tools

Implements a progressive 4-phase loop to refine code context retrieval and solve the subagent context discovery problem.

This skill provides a standardized pattern for AI agents to discover relevant code and project context through an iterative four-stage process: Dispatch, Evaluate, Refine, and Loop. By moving away from 'all-or-nothing' context loading, it enables subagents to intelligently identify high-relevance files, learn project-specific terminology, and fill information gaps without exceeding context limits. It is particularly effective for complex software repositories where initial search queries might miss critical implementation details due to unique naming conventions or architectural patterns.

Key Features

01Relevance Scoring System (0.0 to 1.0) for context quality

022 GitHub stars

03Automated terminology and pattern discovery

04Context gap identification to drive subsequent searches

05Smart exclusion of irrelevant file paths to save tokens

064-Phase Iterative Loop (Dispatch, Evaluate, Refine, Loop)

Use Cases

01Implementing new features that require understanding existing architectural patterns

02Bug fixing in large codebases with unfamiliar naming conventions

03Orchestrating subagents that need specific context to perform specialized tasks

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add ysyecust/everything-claude-code iterative-retrieval

For use in Claude.ai and ChatGPT

Key Features

01Relevance Scoring System (0.0 to 1.0) for context quality

022 GitHub stars

03Automated terminology and pattern discovery

04Context gap identification to drive subsequent searches

05Smart exclusion of irrelevant file paths to save tokens

064-Phase Iterative Loop (Dispatch, Evaluate, Refine, Loop)

Use Cases

01Implementing new features that require understanding existing architectural patterns

02Bug fixing in large codebases with unfamiliar naming conventions

03Orchestrating subagents that need specific context to perform specialized tasks

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add ysyecust/everything-claude-code iterative-retrieval

For use in Claude.ai and ChatGPT