Why is the loop limited to 3 cycles?

Limiting to 3 cycles ensures a balance between finding the perfect context and maintaining performance, preventing the agent from getting stuck in infinite search loops.

How does the evaluation scoring work?

Files are scored from 0.0 to 1.0: High (0.8+) for direct implementation, Medium (0.5-0.7) for related patterns, and Low (below 0.5) for indirect or irrelevant content.

What is the subagent context problem?

It is a challenge where specialized AI agents lack the initial project knowledge to know which files or patterns are needed to complete a task, leading to incomplete or inaccurate work.

Can this skill help with unfamiliar codebases?

Yes, it is specifically designed to discover 'tribal knowledge' and unique project terminology by learning from the results of its initial broad searches.

Iterative Context Retrieval

Name: Iterative Context Retrieval
Author: oabdelmaksoud

byoabdelmaksoud

0•

개발자 도구

Optimizes AI context gathering by progressively refining file searches through a structured four-stage evaluation loop.

The Iterative Retrieval skill solves the common 'subagent context problem' where AI agents struggle to identify necessary project files without prior knowledge of the codebase. By implementing a systematic cycle of Dispatch, Evaluate, Refine, and Loop, this skill enables Claude to start with broad queries and narrow down to high-relevance code (scoring 0.7 or higher). This pattern prevents context window bloat while ensuring the agent discovers project-specific terminology and hidden dependencies that traditional one-shot searches often miss.

주요 기능

01Automated identification of context gaps and missing terminology

020 GitHub stars

03Quantitative relevance scoring (0.0 to 1.0) for retrieved content

04Optimized context management limited to a maximum of 3 refinement cycles

05Four-stage retrieval cycle: Dispatch, Evaluate, Refine, and Loop

06Dynamic search criteria refinement based on initial findings

사용 사례

01Fixing bugs in large codebases where the relevant logic is spread across multiple unknown files

02Implementing new features that require alignment with existing middleware and architectural patterns

03Onboarding subagents to complex projects with unique naming conventions and internal terminology

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

01Automated identification of context gaps and missing terminology

020 GitHub stars

03Quantitative relevance scoring (0.0 to 1.0) for retrieved content

04Optimized context management limited to a maximum of 3 refinement cycles

05Four-stage retrieval cycle: Dispatch, Evaluate, Refine, and Loop

06Dynamic search criteria refinement based on initial findings

사용 사례

01Fixing bugs in large codebases where the relevant logic is spread across multiple unknown files

02Implementing new features that require alignment with existing middleware and architectural patterns

03Onboarding subagents to complex projects with unique naming conventions and internal terminology

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add oabdelmaksoud/superclaw iterative-retrieval

For use in Claude.ai and ChatGPT