research

Orchestrates multi-agent deep research with iterative review loops to produce sourced reports.

Updated Mar 18, 2026
One-click install
npx skills add https://github.com/pgoell/pgoell-claude-tools --skill research-pgoell
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: research
Source: https://github.com/pgoell/pgoell-claude-tools/tree/main/plugins/research/skills/research
Command: npx skills add https://github.com/pgoell/pgoell-claude-tools --skill research-pgoell

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Answering complex questions thoroughly requires gathering evidence from many sources, reconciling contradictions, and writing a defensible report, which is slow and error-prone when done ad hoc. This Skill automates that entire pipeline with parallel research agents and independent review loops. ## Core Features & Use Cases - Parallel deep research: Decomposes a brief into topic clusters and dispatches researcher agents that search iteratively until evidence saturation. - Synthesis with review loops: A synthesis agent builds a thesis and evidence map, then an independent reviewer flags coverage, evidence, source-quality, logic, and structure issues until it passes. - Report writing with review: A writer agent renders the approved synthesis into a polished report (HTML by default, Markdown on request), reviewed for accuracy, format, prose, and flow. - Use Case: Ask for a deep investigation into a market trend or technical topic and receive a structured report with inline citations, credibility tags, a methodology section, and a full source list. ## Quick Start Research the current state of remote work productivity evidence and produce a sourced report.

Frequently Asked Questions about research

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

FAQPage Schema
How do I run a deep research investigation with sources?

State your topic, scope, and audience, and the skill plans topic clusters, dispatches parallel researcher agents that search until evidence saturation, synthesizes findings, and writes a cited report. Output defaults to an HTML report under reports/<topic>-<date>/.

What output formats does the research report support?

The default output is a self-contained HTML report following the bundled template. You can override per invocation by asking for a Markdown report, in which case the deep-mode Markdown template structure is used.

Can this research skill run without subagent support?

Yes. When the host platform cannot dispatch subagents, each phase (research, synthesis, review, writing) runs inline in the orchestrator with the same artifact boundaries. The skill warns you about the increased context and runtime cost.

How does the skill ensure research report quality?

Independent reviewer agents check the synthesis for coverage, evidence gaps, source quality, logic, and structure, then check the report for accuracy, format, prose, and flow. Loops repeat until a PASS verdict, with stall detection and user check-ins every three iterations.

When should I not use the deep research workflow?

Skip it for simple factual questions answerable from existing knowledge; the multi-agent pipeline is designed for complex, multi-source investigations. The skill itself asks whether you want a thorough investigation or a direct answer when intent is ambiguous.