# Deep Research Stanford Storm

> Run Stanford STORM for perspective-driven, article-style deep research reports with retriever-backed citations. Use when a task needs Wikipedia-like synthesis, outline-driven research, or comparison against heavier open-source research frameworks.

- Skill: `closedloop-technologies/deep-research-stanford-storm` (Agent Skill, multi-file: 5 files)
- Install (CLI): `npx skillmds@latest add closedloop-technologies/deep-research-stanford-storm`
- Raw SKILL.md: https://api.skillmd.com/api/skills/closedloop-technologies/deep-research-stanford-storm/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Research & Search
- Author: closedloop-technologies (https://skillmd.com/u/closedloop-technologies)
- Updated: 2026-09-10
- Page: https://skillmd.com/skills/closedloop-technologies/deep-research-stanford-storm

---


# Stanford STORM

Use this skill for Wikipedia-style article generation and perspective-driven
research. STORM is usually slower and heavier than API search tools, so reserve
it for article synthesis benchmarks.

## Environment

```bash
op run --env-file .env.climbhill -- python .agents/skills/deep-research-stanford-storm/scripts/run_storm.py \
  --topic "AI for Science research agents" \
  --rm-name you \
  --fast-model gpt-4o-mini \
  --strong-model gpt-4o-mini
```

Common 1Password-backed variables:

- `OPENAI_API_KEY` or another LiteLLM-compatible model key
- `BING_SEARCH_API_KEY` or `YDC_API_KEY`

## Cost Controls

- Use the same economical model for fast and strong model settings in smoke
  tests.
- Prefer narrow topics.
- Watch intermediate output volume because STORM can generate many files.

## Source Link

- https://github.com/stanford-oval/storm