Can I resume a research session if it's interrupted?

Yes, the skill includes persistent session management. You can use commands like '/research list' to see previous sessions and '/research resume' to pick up exactly where an agent left off.

How does the skill handle complex reasoning vs. simple data gathering?

It uses Smart Model Routing to assign tiers: 'scientist-low' (Haiku) for data gathering, 'scientist' (Sonnet) for analysis, and 'scientist-high' (Opus) for complex architectural reasoning.

What is AUTO mode in the research skill?

AUTO mode allows the research skill to run the entire multi-stage workflow—from initial decomposition to final report generation—autonomously without requiring manual user checkpoints.

Is there a limit to how many agents run at once?

Yes, to prevent resource exhaustion and ensure stability, the skill has a concurrency limit of 5 parallel scientist agents at any given time.

How does the agent ensure the research findings are accurate?

The skill includes a mandatory Verification Loop stage where an agent cross-validates findings from all parallel stages to check for contradictions, gaps in coverage, and evidence quality.

Research Multi-Agent Orchestrator

Name: Research Multi-Agent Orchestrator
Author: Yeachan-Heo

byYeachan-Heo

•

3,720

데이터 과학 및 ML

Orchestrates parallel scientist agents to perform deep, multi-stage research and synthesis within codebases and documentation.

소개

The Research skill automates complex investigative tasks by decomposing a single research goal into multiple parallel stages. It utilizes specialized scientist agents with varying intelligence tiers—Haiku for data gathering, Sonnet for standard analysis, and Opus for complex reasoning—to conduct investigations simultaneously. The skill autonomously cross-validates findings to ensure consistency and generates a structured research report complete with evidence tags and visualizations, making it an essential tool for large-scale security audits, architectural reviews, or deep-dive performance analysis.

주요 기능

Comprehensive report generation featuring evidence-backed findings and architecture diagrams
3,720 GitHub stars
Autonomous decomposition of research goals into 3-7 independent parallel stages
Automated verification loop to cross-validate findings and identify conflicts
Structured session management with the ability to pause, resume, and track progress
Smart model routing that matches task complexity to specific Claude model tiers

사용 사례

Conducting comprehensive security audits and identifying vulnerability patterns across a project
Performing deep architectural analysis to document flows in large or unfamiliar codebases
Comparing performance characteristics and trade-offs between different technical implementations

소개

주요 기능

Comprehensive report generation featuring evidence-backed findings and architecture diagrams
3,720 GitHub stars
Autonomous decomposition of research goals into 3-7 independent parallel stages
Automated verification loop to cross-validate findings and identify conflicts
Structured session management with the ability to pause, resume, and track progress
Smart model routing that matches task complexity to specific Claude model tiers

사용 사례

Conducting comprehensive security audits and identifying vulnerability patterns across a project
Performing deep architectural analysis to document flows in large or unfamiliar codebases
Comparing performance characteristics and trade-offs between different technical implementations