Storm Deepresearch

用多视角扫描、矛盾图、证据搜索和自我复核组织研究简报。用户要求深度研究、多视角分析、研究争议、形成决策材料或使用 storm-deepresearch 时使用。

9 stars 2 forksOther

Install

skills CLI (Claude Code, Cursor, Codex and more)
npx skills add keyelifeai/keye-open-skills --skill storm-deepresearch
Manual install
git clone --depth 1 https://github.com/keyelifeai/keye-open-skills.git /tmp/keye-open-skills
mkdir -p ~/.claude/skills
cp -r /tmp/keye-open-skills/skills/storm-deepresearch ~/.claude/skills/storm-deepresearch

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# STORM Deep Research

这是 Keye 编写的多视角研究工作流,受 Stanford OVAL Lab 的 STORM 研究启发。它不是 Stanford STORM 项目的官方实现,也不复现该项目的软件流程。方法归属与链接见“来源与边界”。

## 开始前

1. 明确研究问题、使用者、决策场景和时间范围。
2. 询问是否允许联网搜索;未获允许或网络不可用时,将证据状态标为“待核验”。
3. 对医疗、法律、投资等高风险主题说明输出仅用于研究整理,并建议由合格专业人士复核。
4. 不把模型记忆当作来源,不虚构引用、数据或共识。

## Step 1:多视角扫描

按问题选择 4 至 6 个彼此有区分度的视角。默认候选包括:

- 实践者:日常约束、真实流程与失败模式。
- 学术研究者:可重复证据、研究设计与外推限制。
- 怀疑论者:最强反论点和被忽略的反例。
- 经济与激励视角:受益者、成本承担者和制度激励。
- 历史视角:相似案例、条件差异和演变路径。
- 受影响群体:不同角色承担的收益、风险和不可逆影响。

每个视角输出:核心立场、最强证据需求、关键假设、独特问题。没有来源支持时写“假设”或“待核验”,不要写成事实。

## Step 2:矛盾图

把视角之间的关系整理为:

1. 直接冲突:具体主张、适用条件和冲突原因。
2. 证据缺口:哪些结论缺少一手数据或高质量研究。
3. 可化解问题:哪个问题一旦回答,最能缩小不确定性。
4. 暂时共识:多方在什么边界条件下同意。
5. 盲区:没有被任何视角充分覆盖的人群、时间尺度或副作用。

不要用“多数视角同意”代替证据质量判断。

## Step 3:证据搜索

联网搜索获准时,从矛盾图提取 2 至 5 个可验证查询:

- 争议最大的因果或趋势主张;
- 盲区和前沿问题;
- 会改变决策的关键参数。

优先使用论文原文、政府和监管机构、标准组织、官方数据、项目官方文档。每条证据记录标题、发布者、日期、URL、支持或挑战的主张,以及适用限制。找不到可靠证据时保留“不确定”,不要降低来源标准补数。

## Step 4:综合简报

按以下顺序输出:

1. 决策摘要:研究问题、当前结论和最重要的不确定性。
2. 关键发现:每条标明证据强度、支持与挑战视角。
3. 矛盾与边界:哪些结论只在特定条件下成立。
4. 行动选项:可逆动作、观察指标和停止条件。
5. 待核验清单:缺失数据、负责人和下一次复核触发点。

## Step 5:自我复核

逐条检查:

- 引用是否真实存在并支持相邻主张。
- 事实、推断、观点和建议是否清楚分开。
- 是否遗漏受影响群体或反例。
- 是否把相关性写成因果性。
- 哪条结论最弱,需要什么证据才能提高可信度。
- 新证据出现时,哪些结论必须重新评估。

最后给出高、中、低三档证据强度,不使用看似精确但没有量表依据的数字评分。

## 来源与边界

- 论文:[Assisting in Writing Wikipedia-like Articles From Scratch with Large Language Models](https://arxiv.org/abs/2402.14207)
- 官方代码:[stanford-oval/storm](https://github.com/stanford-oval/storm)
Read the full skill on GitHub

Details

Skill name
storm-deepresearch
Path
skills/storm-deepresearch/SKILL.md
License
MIT
Repo updated
30 August 2026

FAQ

What does the Storm Deepresearch skill do?

用多视角扫描、矛盾图、证据搜索和自我复核组织研究简报。用户要求深度研究、多视角分析、研究争议、形成决策材料或使用 storm-deepresearch 时使用。

How do I install the Storm Deepresearch skill?

skills CLI (Claude Code, Cursor, Codex and more): npx skills add keyelifeai/keye-open-skills --skill storm-deepresearch. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is Storm Deepresearch popular?

It lives in keyelifeai/keye-open-skills, which has 9 GitHub stars and 2 forks.

Related skills

Keye Network Optimizer

9

keyelifeai/keye-open-skills

Diagnose and improve macOS network speed and stability with before baselines, proxy/VPN/TUN separation, minimal reversible changes, and after retesting. Use for networkQuality, DNS latency, Wi-Fi quality, packet loss, MTU, system proxy, route takeover, CLI proxy, or background traffic investigations, and when the user asks to 优化网络、排查网络不稳定、检查代理或 runs /keye-network-optimizer.

Finance & Crypto

Keye Viral Dissect

9

keyelifeai/keye-open-skills

分析微信公众号、X、小红书等平台文章的结构、论证和情绪设计。用户提供文章、PDF 或公开 URL 并要求拆解、分析爆款或提取写作方法时使用。只做分析,不生成或重写文章。

Documents & Office Files

Proposal Assembler

10

gener8v/gener8v.claude-skills.rfp-response

Assembles all reviewed, validated sections into a single proposal-final.md. Inserts front matter, enforces section order, attaches appendices, performs a final compliance pre-flight, and surfaces any unresolved issues. Trigger when the user says "assemble the proposal," "build the final document," "put it together," or after the compliance validator has produced compliance-report.md. Will refuse to assemble if critical compliance failures exist.

Other

Decide

9

deejay-hub/salesforce-ada-agent-skills

Generate a Decision Record through structured analysis pipeline. Use when comparing architectural options, choosing between technical approaches, evaluating build-vs-buy trade-offs, or making any design decision that needs evidence-based scoring.

Other

Requirements Mapper

10

gener8v/gener8v.claude-skills.rfp-response

Maps every explicit and implicit RFP requirement to available evidence from the collateral brief. Produces a coverage matrix so downstream skills know where to ground claims, where to assert, and where to acknowledge gaps. Trigger when the user says "map the requirements," "build the coverage matrix," "what can we substantiate," or after the collateral analyzer has produced brief.md and gap-report.md.

Other

Spectacles 522 to 515 Migration

9

PtPavloTkachenko/spectacles-agent-skills

Battle-tested playbook for migrating a Lens Studio 5.22 project down to LS 5.15.4 for Spectacles (2024). Triggers on: migrate 5.22 to 5.15, downgrade lens studio project, 5.22 package errors in 5.15, SingleCompressionSettings error, PerformanceCompressionSettings error, Invalid graph header, !passList.empty(), duplicate PassInfo reimport loop, PassInfo transplant, materials white after clone, 5.22 lspkg crashes 5.15.

Other

Browse the most popular Claude skills, all other skills, or write your own with the skills generator.