How to use the Deep Research Methodology agent skill

What the skill makes the agent do

The skill follows a five-phase research process: Define the question and scope, Search across diverse sources, Evaluate source quality and limitations, Synthesize findings into a coherent answer, and Verify critical claims while listing remaining uncertainties.

It also gives the agent a source-evaluation checklist called CRAAP (Currency, Relevance, Authority, Accuracy, Purpose), a scoring scale from A to F, and a four-level verification scheme. Claims can be marked as reported by a single source, confirmed by multiple sources, verified against a primary source, or widely accepted.

• Breaks complex questions into sub-questions
• Runs at least three search strategies per question, including direct, authoritative, and contrarian queries
• Scores sources with CRAAP and labels confidence levels
• Provides report templates and citation formats (inline URL, footnotes, APA, numbered references)

Install the skill (ZIP download or Claude upload)

Download the ZIP from the skill's page on this site and unzip it. The package contains two files: HAND.toml and SKILL.md.

To make the skill available to your own Claude Code setup, unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder to share it with one project.

In the Claude apps you can instead upload the skill under Settings → Capabilities → Skills.

Trigger the skill with natural requests

You do not need a special command. Ask for research in plain language and say that you want a careful, multi-source approach, so the agent applies the method rather than answering from memory.

The skill's own examples show the kinds of questions it handles: factual, comparative, causal, predictive, how-to, survey, and controversial.

• Compare React and Vue for large apps, using several sources and labelling what is corroborated.
• Research whether remote work is more productive, presenting multiple perspectives and primary sources.
• Map the options for vector databases and note the limits of each source.
• Help me decide whether our startup should use microservices. Break the question into sub-questions first.

What the skill needs and what it does not do

The skill lists no requirements. It runs no scripts and uses no network access. It is prompt-only guidance.

That also means it cannot search for anything. It contains no search or web-fetch code, so for live results the agent needs a separate search tool. The skill shapes how the agent searches and judges sources; it does not find them or confirm that they are current.

• Requires: nothing listed
• Runs scripts: no
• Uses network: no
• Connects to databases or adds tools: no

Safety notes for medical, legal and financial questions

The skill only shapes method. It does not verify facts or guarantee that sources are current, so its output should not be treated as authoritative for medical, legal or financial decisions.

Its own guidance includes checking currency, which means anything over two years old may be outdated for technology topics. Check dates and primary sources yourself before acting on findings.

Tips for better research results

Name the audience and the decision you are making, because the skill's relevance check depends on it. Ask for the output format you need, such as the Brief Report or Detailed Report template.

Ask the agent to label each claim with its verification level, so you can see which findings rest on a single source.

• Give the question a clear scope, including the time period you care about
• Request a Caveats or gap analysis section for anything you will rely on
• Use the citation format your course or publication requires