Decompose large-context tasks into partitions and spawn parallel subagents.
npx skills add https://github.com/belumume/claude-skills --skill rlm-orchestrator
Please help me install this Agent Skill. Skill: rlm-orchestrator Source: https://github.com/belumume/claude-skills/tree/main/rlm-orchestrator Command: npx skills add https://github.com/belumume/claude-skills --skill rlm-orchestrator
This skill orchestrates large-context tasks by decomposing them into independent partitions, spawning parallel subagents, and aggregating results to produce a unified outcome. It addresses context window limitations and complex multi-file or multi-source workloads.
Start by describing a large, multipart task and specify partition boundaries, then invoke the orchestrator (e.g., mention RLM or /rlm-orchestrator). Monitor batch results and review aggregated outcomes to guide next steps.
High-intent search queries and answers about installing and using this skill.
To analyze large codebases exceeding a single context window, you can partition the task into independent parts and dispatch parallel subagents. This orchestration approach spawns fresh-context agents concurrently and aggregates their structured summaries into a unified outcome.
The best way to run parallel subagents for multi-file analysis is using a partition, spawn, and aggregate protocol. It decomposes workloads into independent partitions, executes up to ten concurrent subagents, and aggregates their results iteratively to resolve gaps.
Orchestrate subagents to synthesize research documents by defining partition boundaries for the workload, invoking the orchestrator, and monitoring batch results. The system processes partitions concurrently and returns structured summaries to guide iterative refinement until completion.
Yes, Claude Code supports spawning parallel subagents with fresh context. This workflow is designed specifically for Claude Code tooling environments where child agents cannot be spawned in a single step, enabling concurrent execution and result aggregation.
Iterative refinement during parallel execution works by identifying gaps in aggregated subagent results and spawning follow-up batches. This cycle repeats until completion, ensuring the unified outcome addresses all partition requirements without context window limitations.
You should not use parallel subagent orchestration for tasks that cannot be decomposed into independent partitions. The workflow requires Claude Code tooling and relies on partition boundaries, making it unsuitable for single-context tasks or non-decomposable sequential workloads.
ruskibeatsDecompose large tasks into actionable subagent tasks for parallel dispatch in Clanker Ops.
derekhu0002Refine codebase context for subagents through a four-phase retrieval loop.
shakestzdOrchestrate parallel agent workflows from analysis to aggregation.
leonlettoPartition large research workloads across parallel sub-agents and consolidate findings into a report.
