Can RLM Mode perform batch refactoring?

Yes, the rlm-batch command is specifically designed to apply transformations, updates, or generations across multiple files concurrently using parallel workers.

What is RLM Mode in Claude Code?

RLM Mode stands for Recursive Language Model mode. It is a processing strategy that decomposes large-scale operations into smaller chunks, allowing Claude to process massive repositories that would otherwise exceed its context window.

Is it possible to filter which files RLM processes?

Absolutely. RLM commands support glob patterns (e.g., **/*.ts) and path extraction to target specific directories or file types accurately.

When should I use RLM Mode instead of standard commands?

You should use RLM Mode for tasks involving the entire codebase, recursive directory scans, or when processing more than 20 files at once to avoid context overflow and quality degradation.

How does RLM handle context window limits?

Instead of loading all files into a single prompt, RLM processes files in independent batches and then aggregates those results into a final coherent summary or report.

RLM Recursive Codebase Architect

Name: RLM Recursive Codebase Architect
Author: jmagly

byjmagly

•

102

•

生产力与工作流

Decomposes large-scale codebase operations into recursive tasks to process massive repositories without context window limits.

The RLM Mode skill solves the context window constraints of AI agents by implementing Recursive Language Model (RLM) processing. It intelligently identifies project-wide requests—such as auditing every file, batch refactoring hundreds of components, or summarizing complex architectures—and breaks them into manageable chunks. By fanning out tasks across a repository and aggregating the results, this skill ensures high-quality output and reliable performance on enterprise-scale projects that would otherwise exceed model limits or cause performance degradation.

主要功能

01Hierarchical Summarization (rlm-summarize) for mapping complex repository structures

02Configurable parallel processing for high-performance multi-file operations

03Intelligent scope detection and automatic RLM suggestions based on file volume

04102 GitHub stars

05Batch Processing (rlm-batch) for large-scale automated refactoring and transformations

06Recursive Query (rlm-query) for exhaustive project-wide information extraction

使用场景

01Auditing an entire enterprise repository for security vulnerabilities or deprecated API patterns

02Executing massive refactoring tasks across hundreds of files with parallel worker support

03Generating deep architectural overviews of nested directory structures to understand legacy codebases

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jmagly/aiwg rlm-mode

For use in Claude.ai and ChatGPT