storm-research

Generates multi-perspective, citation-verified HTML research briefings using five expert lenses and adversarial peer review.

2|1|Updated Feb 12, 2026
One-click install
npx skills add https://github.com/cilerler/lillian --skill storm-research-cilerler
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: storm-research
Source: https://github.com/cilerler/lillian/tree/main/plugins/ai-toolkit/skills/storm-research
Command: npx skills add https://github.com/cilerler/lillian --skill storm-research-cilerler

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes assets (resource) components.

What problem does it solve? Single-source research summaries often miss conflicting viewpoints and contain unverified or fabricated claims. This Skill turns one topic into a fact-checked, multi-perspective HTML briefing where every citation is independently verified against its primary source before delivery. ## Core Features & Use Cases - Five Expert Lenses: Runs parallel research subagents as Practitioner, Academic, Skeptic, Economist, and Historian, then maps where they contradict and agree. - Citation Verification Pipeline: Adversarially peer-reviews its own draft, verifies every citation against primary sources, and tags each as Confirmed, Corrected, Contested, Demoted, or Fabricated. - Self-Contained HTML Report: Produces a single styled HTML file with reliability-scored findings, a hidden-connection analysis, a claim safety guide, and a frontier question. - Use Case: Ask for a STORM briefing on a contested topic like AI coding assistant productivity, and receive a verified report showing what practitioners, academics, and skeptics each found, which claims survived fact-checking, and what is safe to assert. ## Quick Start Run a storm research briefing on the topic of your choice and deliver the verified HTML report.

Frequently Asked Questions about storm-research

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

FAQPage Schema
How do I run a STORM research briefing on a topic?

Invoke the skill with a topic, for example by asking for a storm report on that subject. It runs a five-phase pipeline: scoping, five parallel expert-lens research agents, contradiction mapping, HTML report synthesis, and adversarial citation verification before delivering the file.

What does the multi-perspective research report contain?

The self-contained HTML report includes a 60-second summary, five reliability-scored key findings, a hidden-connection analysis, a missing-sixth-lens blind spot, actionable recommendations, a claim safety guide, a frontier question, and a reference list with per-citation verification status.

Can storm research run without web access?

No. Web access is a hard requirement because every lens and citation must trace to a real fetched source. If the platform cannot fetch the web, the skill stops at Phase 0 rather than generating a report from training memory.

When should I not use storm research?

Skip it for simple factual lookups, since the pipeline spawns roughly 9-11 subagents per run and is deliberately heavyweight. It is designed for contested topics where multiple viewpoints and verified claims justify the cost.

How are citations verified in the report?

Verification subagents fetch each primary source directly and confirm or correct the title, authors, figures, and peer-review status. Verdicts map to status tags: Confirmed, Corrected, Contested, Demoted, or Fabricated, and the report banner tallies the results.