rlm

Coordinate sub-agents to analyze large codebases without overflowing context.

1|Updated Feb 2, 2026

Search Tags:#codebase #parallel #analyze #recursive #map-reduce #large-repo

Ready to run

npx skills add https://github.com/micaelmalta/agentic --skill rlm-micaelmalta

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: rlm
Source: https://github.com/micaelmalta/agentic/tree/main/skills/rlm
Command: npx skills add https://github.com/micaelmalta/agentic --skill rlm-micaelmalta

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Large codebases pose memory and context challenges. The RLM skill coordinates sub-agents to read, map, and synthesize insights without loading the entire project into a single context, enabling scalable analysis.

Core Features & Use Cases

Recursive planning and parallel mapping to extract cross-file patterns, API usages, and architecture signals from 100+ files.
Phase-driven workflow: choose engine, index & filter, parallel map, then reduce & synthesize results.
Use Case: analyze a monorepo to locate all occurrences of a deprecated API and summarize usage patterns across modules.

Quick Start

Activate the RLM workflow to scan a large repository and generate a concise summary of key findings.

Frequently Asked Questions about rlm

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I analyze a large codebase without overflowing the main context window?▼

To analyze a large codebase without context overflow, you can use recursive map-reduce orchestration to coordinate sub-agents that read and synthesize code in parallel, keeping the main context clear.

What is the best way to find deprecated API usage patterns across a monorepo?▼

The best way to find deprecated API usage across a monorepo is to run a parallel mapping workflow that scans hundreds of files simultaneously and aggregates cross-file patterns into a single summary.

Can I use parallel agents to extract architecture signals from multi-language projects?▼

Yes, you can use parallel agents to extract architecture signals from multi-language projects by spawning sub-agents for selective data extraction and chunked processing, then recursively synthesizing the aggregated results.

How do I process 100 or more files in a repository without loading them all at once?▼

You can process 100+ files in a repository by applying a phase-driven map-reduce workflow that indexes and filters files first, then dispatches parallel sub-agents for chunked processing without loading everything at once.

Does recursive codebase analysis work with strict mode orchestration engines?▼

Yes, recursive codebase analysis supports both native and strict mode orchestration engines, allowing you to spawn parallel tasks and aggregate results while maintaining strict execution constraints.

What are the limitations of using sub-agents for large repo analysis?▼

Limitations of using sub-agents for large repo analysis include the need for engines that support parallel task spawning and recursive synthesis, as standard single-context processing cannot handle 100+ files without overflowing memory.