skills.rest
kellymears
kellymearsCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

research

Conduct multi-source web research and generate cited reports with verification-tagged claims.

Updated Feb 6, 2026
Search Tags:#pdf#research#web search#report generation#citations#anti-hallucination#fact checking
One-click install
npx skills add https://github.com/kellymears/agents --skill research-kellymears
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: research
Source: https://github.com/kellymears/agents/tree/main/plugins/research/skills/research
Command: npx skills add https://github.com/kellymears/agents --skill research-kellymears

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires reportlab, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve? Research reports built on fabricated citations or unverified claims can cause real harm — bad medical decisions, flawed legal strategies, wasted investments. This Skill enforces an anti-hallucination discipline: every claim is tagged with a verification status, gaps are disclosed instead of filled, and conflicting sources are presented honestly. ## Core Features & Use Cases - Structured six-phase workflow: scope and decompose the question, search and gather sources, analyze perspectives, optionally run quantitative checks, compose the report, and deliver it. - Verification-tagged citations: every claim is labeled Verified, Retrieved, Training Data, or Unverifiable, with inline confidence tags for medical, legal, and financial topics. - Qualitative confidence framework: five-dimension self-assessment (source quality, consensus, recency, domain familiarity, evidence strength) plus plausibility tiers for competing perspectives. - Multiple output formats: terminal tree-style briefs, markdown files, or styled PDF reports generated via a reportlab-based script. - Use Case: Ask "is intermittent fasting actually supported by evidence?" and receive a report with competing viewpoints, plausibility tiers, cross-referenced statistics, and a citation list distinguishing verified sources from training-data recall. ## Quick Start Ask the agent to research a topic such as "do a deep dive on the evidence for intermittent fasting" and confirm the proposed research plan when prompted.

Frequently Asked Questions about research

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

FAQPage Schema
How do I get a research report with verified citations?▼

Ask the agent to research your topic or fact check a claim. It proposes a research plan with sub-questions, waits for your confirmation, then searches and fetches sources, tagging every claim as Verified, Retrieved, Training Data, or Unverifiable.

How does the skill prevent fabricated sources and hallucinated citations?▼

An anti-hallucination contract forbids inventing URLs, authors, titles, or dates. Every claim carries a verification status, information gaps are disclosed explicitly, conflicting sources are presented side by side, and a self-check list runs before delivery.

What output formats does the research report support?▼

Reports can be delivered as a terminal tree-style brief for short results, a markdown file written to the working directory, or a styled PDF generated from the markdown using the bundled report-pdf.py script built on reportlab.

Can the research skill verify numerical or statistical claims?▼

Yes. A conditional quantitative phase runs Python scripts to cross-reference numbers across sources, calculate derived statistics, and model scenarios. The script and its output are included in the report so results are reproducible.

What are the limitations of AI-generated research reports?▼

Coverage depends on what web search returns; paywalls, thin results, and very recent topics limit source quality. Training-data claims are legitimate but unverified in-session, so the report tags them and notes scope limitations honestly.

Related Skills

Explore More Education & Research Skills→
mmnavarrmmnavarr

research

Delegates topic research to a background agent that cites primary sources in Markdown notes.

Community
Basic
analystacademyanalystacademy
35

content

Converts source material into three to four finished LinkedIn post drafts.

Community
Advanced
FlintmereFlintmere

writer

Generate cited long-form prose from briefs and reference sources.

Official
Advanced
kortix-aikortix-ai
20.2k

deep-research

Conducts multi-source web research and produces cited, evidence-based reports.

Official
Advanced