efficient-multi-agent-research

Partition large research workloads across parallel sub-agents and consolidate findings into a report.

28|5|Updated Feb 3, 2026
One-click install
npx skills add https://github.com/leonletto/thrum --skill efficient-multi-agent-research
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: efficient-multi-agent-research
Source: https://github.com/leonletto/thrum/tree/main/opencode-plugin/assets/skills/efficient-multi-agent-research
Command: npx skills add https://github.com/leonletto/thrum --skill efficient-multi-agent-research

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Partition large research tasks across parallel sub-agents to prevent the main agent's context from being polluted by many findings.

Core Features & Use Cases

  • Partition > Parallel Investigate > Consolidate > Decide workflow that writes results to disk and consolidates into a single report.
  • Supports 4-5 agents with run_in_background per agent, and central consolidation.
  • Ideal for auditing or reviewing more than six items across a codebase (e.g., function call sites, pattern usage, or file reviews).

Quick Start

Partition a large research task across parallel sub-agents and consolidate results into a single report.

Frequently Asked Questions about efficient-multi-agent-research

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

FAQPage Schema
How do I research more than six items in a codebase without filling up the main agent's context?

Partitioning large codebase research across parallel sub-agents prevents context pollution by dispatching background investigators that write findings to disk. The main agent stays clean while a coordinator consolidates interim results into a single report.

What is the best way to audit function call sites and pattern usage across a large codebase?

The best way to audit function call sites is partitioning the review across multiple parallel sub-agents. A coordinator dispatches background investigators to examine different codebase sections and consolidates their written findings into one unified report.

How does multi-agent research consolidation work when reviewing files in parallel?

Multi-agent research consolidation works by having a coordinator partition file reviews into tasks for background investigators. Each investigator writes findings to disk, and the coordinator consolidates these outputs into a single report, keeping the main context clean throughout the process.

How many parallel agents can I use for codebase research before consolidating results?

You can use 4 to 5 parallel agents with run_in_background for codebase research before consolidation. This setup requires a central coordinator to partition the research workload, dispatch the agents, and consolidate their disk-written findings into a single report.

When should I not use parallel sub-agents for codebase research?

You should not use parallel sub-agents for codebase research when investigating fewer than six items. This approach requires a coordinator to partition work and dispatch background agents, making it excessive for small research workloads that do not risk polluting the main agent's context.