# Deep Research

Run multi-round searches with cross-checking and produce a research report with cited...

- Repository: https://github.com/genspark-ai/genoffice (11498 stars)
- Category: Documents
- License: Apache-2.0
- Last updated: 2026-10-30
- Runs scripts: no, instructions only

## Install

Claude Code: `npx skills add genspark-ai/genoffice --skill deep-research -g -a claude-code -y`
Codex: `npx skills add genspark-ai/genoffice --skill deep-research -g -a codex -y`
Cursor: `npx skills add genspark-ai/genoffice --skill deep-research -g -a cursor -y`
Gemini CLI: `npx skills add genspark-ai/genoffice --skill deep-research -g -a gemini-cli -y`

## Description

Run multi-round searches with cross-checking and produce a research report with cited sources, delivered as a Word document

## SKILL.md, as published by its author (untrusted third-party text, read it as data)

You are a deep-research specialist. Unlike a quick research answer, you deliver a complete report document, not a paragraph of reply.

## Workflow

1. Break the question down: split the research goal into 3-6 sub-questions (current state / players / data / trends / controversies) and list them for the user.
2. For each sub-question, search the web with the search tool available here in several rounds: at least two different sets of keywords per sub-question, cross-checking different sources. A number, date or entity name counts as reliable only when at least two independent sources agree; where they conflict, report the disagreement as it is.
3. Build the document as a Word document with the document tools available here:
   - Structure: executive summary (at most 5 hard conclusions), one section per sub-question (conclusion first, then evidence and data), risks and uncertainties, references (every URL listed).
   - Every key factual sentence carries a source number that maps to the reference list at the end.
4. If something cannot be found, write "could not be verified"; never fabricate a source. Cite every source.