Produce a cited research report in Claude Code with Deep Research

Who it is for

This skill is for people doing fuller research in Claude Code who need a cited report, source-confidence scores, and automatic checks. Officially it is a research engine for Claude Code: it produces reports backed by citations and includes multiple search sources plus automatic checks.

It is not a search-box substitute and it will not only hand you a few links. A better use is: you already have a clear topic, you are willing to let the assistant finish the research flow, and you accept that the report is written into the local documents directory. If you only want one short question, you do not need to start the full research.

Official notes:

https://github.com/199-biotechnologies/claude-deep-research-skill

First try in 3 minutes

First confirm you are using skills in Claude Code. The official install command is only this one:

git clone https://github.com/199-biotechnologies/claude-deep-research-skill.git ~/.claude/skills/deep-research

This clones the repository into the local skill directory and writes to the filesystem. Before running it, confirm the target path is ~/.claude/skills/deep-research. Do not add npx or another installer. The page says basic use needs no extra dependencies.

After install, trigger one low-risk study with the official example wording. The page says: deep research on the current state of quantum computing. This does external search and may write the report into the local documents directory. Before sending, confirm you accept network access and file writes. Do not change repository code outside the topic first.

Success signal: the assistant starts cited research through this skill, and you can later see a matching folder in the documents directory. The page says reports are saved to ~/Documents/[Topic]Research[Date]/. If there are no citations or that directory does not appear, stop at this step.

Usage examples

Example 1: start a topic study with the fixed wording. The evidence says you can ask in Claude Code with a request like deep research on .... The first time, use the page's quantum-computing-state example, watch whether it produces cited content and whether it explains source confidence. Do not stuff several unrelated topics into the trial run at once.

Example 2: check the saved report. The evidence says the report is written to ~/Documents/[Topic]Research[Date]/. After research ends, open that directory first, confirm files were created, then decide whether to copy conclusions into other documents. Do not assume the study is finished if you have not seen that directory. The materials do not list filenames inside the directory; do not invent the missing pieces.

Limits

This is a skill for Claude Code, and the install path also points at Claude's skills directory. Basic use does not require installing other dependencies. The page may mention optional tools, but the first time do not treat them as required installs.

The research process searches the network and writes a report in the local documents directory. Confirm the topic and the write location first. The materials do not write how to turn off disk writes, and they do not write install paths for other chat products. Do not treat the example topic as the only topic you can study, and do not invent a second install command.