research

Coordinate parallel expert investigations into a review-ready research report.

23|2|Updated Mar 7, 2026

Search Tags:#research #validation #parallel #swarm #synthesis #report #experts

Ready to run

npx skills add https://github.com/kirich1409/krozov-ai-tools --skill research-kirich1409

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: research
Source: https://github.com/kirich1409/krozov-ai-tools/tree/main/plugins/developer-workflow/skills/research
Command: npx skills add https://github.com/kirich1409/krozov-ai-tools --skill research-kirich1409

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Orchestrates parallel domain-expert research before implementation, ensuring diverse perspectives are gathered and structured into a single, review-ready report.

Core Features & Use Cases

Parallel expert tracks: launches up to 5 domain-specific agents (codebase, web, docs, dependencies, architecture) to investigate different angles simultaneously.
Independent data gathering, centralized synthesis: collect findings without bias, then have a separate reviewer (business-analyst) synthesize for completeness and product sense.
Phase-driven workflow with persistence: scoped, tracked research with state persistence to survive context changes and enable recovery.
Artifacts for downstream workflows: produces a final research report and structured state suitable for pipeline integration.

Quick Start

Define your topic and scope, then trigger the swarm to run parallel expert tracks and generate the final research report.

Frequently Asked Questions about research

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

FAQPage Schema

How do I run parallel expert research before starting a software architecture decision?▼

Parallel expert research launches up to 5 domain-specific agents to investigate codebase, web, docs, dependencies, and architecture simultaneously. A separate business-analyst then synthesizes these independent findings into a single, review-ready report.

What is the best way to gather diverse perspectives for a product feasibility study?▼

Gathering diverse perspectives for a product feasibility study requires orchestrating independent data collection across multiple tracks. This approach prevents bias before a centralized reviewer synthesizes the findings into a comprehensive, structured report.

Can I use parallel swarm agents to investigate dependencies and docs simultaneously?▼

Yes, parallel swarm agents can investigate dependencies and docs simultaneously. The swarm coordinates up to 5 domain-specific tracks, including codebase, web, docs, dependencies, and architecture, to illuminate complex topics before implementation.

How does state persistence work during phased expert research workflows?▼

State persistence during phased expert research workflows saves scoped progress to survive context changes. This mechanism enables recovery across research phases, ensuring the final structured state and artifacts are suitable for downstream pipeline integration.

What is needed to trigger a multi-agent research swarm for a complex topic?▼

To trigger a multi-agent research swarm, you must define your topic and scope clearly. The swarm then runs parallel expert tracks across code, web, docs, dependencies, and architecture to generate a final, structured research report.

When should I avoid using parallel expert investigations for architecture decisions?▼

You should avoid parallel expert investigations when a topic lacks sufficient complexity to require multiple domain tracks. The swarm is designed for broad scoping across code, web, docs, and architecture, making it excessive for simple, single-domain validation tasks.