deep-research skill
One /research command builds a cited report with counter-arguments and scenarios in ten minutes, in 19 languages with Russian.
- 615 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The agent browses sites through Scrapling MCP, downloads pages and saves report files to disk.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/deep-researchDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The agent browses sites through Scrapling MCP, downloads pages and saves report files to disk.
This is third-party code. Review the repository files before installing.
What it does
The deep-research skill runs a four-stage pipeline: outline, source collection and verification, chapter writing, final check. The output is a report with conclusions first, tables, references and opposing views. The `-quick`, standard and `-deep` modes finish in 8 to 25 minutes. The topic language is detected automatically, and the README shows requests in Russian, Chinese and others. The skill also reads local PDF, DOCX, TXT and MD files without internet. Pages are fetched by the Scrapling MCP Server. Finished reports open in a local browser with search and PDF/DOCX export.
Who it is for. Analysts, founders and researchers who need a traceable market or technology review instead of a shallow search summary.
Good fit when
- A competitive review of an industry is prepared with tables and sources.
- A policy or technology is analyzed with counter-arguments and scenarios.
- A report is built from a folder of your own PDFs and documents without going online.
- The topic is in Russian and the report must be in Russian.
- An investment memo needs a reference for every number.
Not a fit when
- A one-paragraph answer is enough.
- The report goes to a client without checking the numbers: the skill reduces fabrication risk but does not replace verification.
- Installing Scrapling MCP and running the page scraper on the machine is not possible.
Example request
/research Competitive landscape of AI cloud computing -quickLimitations
Page collection runs through the Scrapling MCP Server, which must be installed and registered separately per its docs. One report launches several agents and spends a noticeable number of tokens. Sources and quality depend on search and site availability, so key figures deserve spot checks. The repository is cross-platform, but how the skill and commands are registered differs per tool and is listed in a README table. The last commit dates from August 2026.
How to disable. Delete the `deep-research` folder from the agent skills directory and remove the Scrapling MCP Server from MCP settings. Saved reports live in the `reports/` folder.
FAQ
How long does a report take?
Quick mode takes 8 to 12 minutes, standard 10 to 15, deep 15 to 25. Modes are chosen with the `-quick` and `-deep` flags.
Does it work offline?
Yes, for local files. MD and TXT are read directly, while PDF and DOCX are parsed after the agent installs pypdf and python-docx.
Where are reports stored?
In the `deep-research/reports/` folder of the skills directory. The `reports-browser/index.html` page shows them in a searchable table.
Which tools are supported?
The README names Claude Code, Codex CLI, Cursor, Windsurf, Cline, OpenCode and others: drop the project into the tool's skill or command folder.
How do I update the skill?
With `/research-update`, which checks for project updates. The version lives in the `VERSION` file.
Discussion
No questions or reviews yet.
Related
SkillsA multi-round research skill: objective, evidence, critique and synthesis go to files, and the report comes after a final check
Feynman — a terminal AI agent for scientific research: literature reviews, source checks, paper critique and comparing a paper with its code
A skill that finds other skills for your task through the DeepSkill Market service and ranks them using reviews from people who ran them
A set of 11 Claude Code skills that audits a third-party paper for self-consistency and fabrication and writes a report.
Autoresearch for skills
autoresearch
A skill that improves another skill: runs it on test inputs, scores outputs with binary checks and mutates the prompt
An autonomous deep-research agent: gathers sources, verifies them and writes a report on any question
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| deep-research (this card) | no | 615 | |
| Deep Research Skills | no | 2,550 | a stepwise research skill; here one run yields a full report in 19 languages and a local report browser |
| Deep Research Agent | no | 371 | a research agent; hoolulu/deep-research stays a skill with a /research command and source verification |
README in short
One run builds a report in an institutional structure: conclusions first, traceable sources, counter-arguments and scenario forecasts.