deep-research · petekp — Claude Code skill: research across sources
deep-research is a Claude Code skill that investigates complex questions across multiple sources and produces cited findings, beyond retrieving a single definition or page.
MIT 196 installs ★ 49 updated 2026-10-25
npx skills add petekp/claude-code-setup --skill deep-research -a claude-codeWhat deep-research does
The agent clarifies what needs investigating, identifies decision criteria, and sets research depth. It may ask for clarification if that would materially change the outcome.
The skill separates primary sources from interpretation, seeks independent confirmation for contested or empirical claims, and follows chains of references. Results organize around findings and disagreements between sources rather than one summary per source. Citations appear next to the claims they support.
Use it when a question spans changing terrain, conflicting statements appear, or you need to verify behavior against your local code. Research continues until requested questions are supported or unresolved gaps are identified.
What it helps with
- investigate a question by consulting multiple sources systematically
- distinguish product facts from interpretation and speculation
- find independent confirmation for contested claims
- organize findings around conclusions, not individual sources
- identify which evidence is unavailable and why
Example prompts for the agent
- Research how Claude Code handles context memory and which versions improved it
- Compare error handling approaches in Cursor and GitHub Copilot using independent reviews
- Check whether our Node 18 project works with Gemini CLI or if a newer version is required
How to install deep-research
1Skills CLI (Vercel)
npx skills add petekp/claude-code-setup --skill deep-research -a claude-code2Manually: put the files in place
~/.claude/skills/deep-research/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/deep-research/
1Skills CLI (Vercel)
npx skills add petekp/claude-code-setup --skill deep-research -a codex2Manually: put the files in place
.agents/skills/deep-research/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add petekp/claude-code-setup --skill deep-research -a cursor2Manually: put the files in place
.cursor/skills/deep-research/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install petekp/claude-code-setup skills/deep-research2Manually: put the files in place
.github/skills/deep-research/SKILL.md1Skills CLI (Vercel)
npx skills add petekp/claude-code-setup --skill deep-research -a gemini-cli2Platform CLI
gemini skills install https://github.com/petekp/claude-code-setup --path skills/deep-researchManually: put the files in place
.devin/skills/deep-research/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add petekp/claude-code-setup --skill deep-research -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about deep-research
How do I install deep-research in Claude Code?
Run npx skills add petekp/claude-code-setup --skill deep-research -a claude-code, then open ~/.claude/skills/deep-research/SKILL.md. For other platforms, replace the -a flag with codex, cursor, gemini-cli, or opencode.
Why use deep-research instead of regular search?
Regular search returns popular pages or the first result. deep-research follows source chains, checks for contradictions, and looks for independent confirmation—essential when one source may be wrong or outdated.
Can deep-research read private code in my project?
Yes, the skill can examine local code to verify how something actually behaves in your environment, rather than relying solely on documentation.
Source text
SKILL.md · MIT
--- name: deep-research description: "Investigate a question that requires sustained research across sources and produce cited findings. Use for deep research requests; skip ordinary lookups and single-document reviews." --- # Deep research Establish the question, decision criteria, and requested depth from the task. Ask only for missing information that would materially change the research. Preserve the user's named topic or model in searches; do not replace an unfamiliar or fast-changing name with a familiar one. Read primary sources and distinguish their claims from your interpretation. Use current local code for questions about the installed implementation. Follow leads and seek independent evidence where a consequential, contested, or empirical claim needs it. One authoritative source may settle a documented product fact; a fixed source count does not establish truth. Use these references only when needed: - [Search strategies](references/search-strategies.md): discovery is stalled, terminology is unclear, or the question spans several kinds of sources. - [Source evaluation](references/source-evaluation.md): sources conflict, evidence is indirect, or a study or benchmark needs scrutiny. - [Reporting](references/reporting.md): producing a substantial comparison or research artifact. Answer with citations near the claims they support. Paraphrase sources, reserve short marked quotes for exact wording that matters, and organize around findings and disagreements rather than one summary per source. Use the host's citation format and quotation limits. Continue until the requested questions are supported or the unresolved gaps are identified. Follow remaining leads that could change the answer within the agreed scope and budget. Report unavailable evidence and limitations; do not imply exhaustive coverage or start unrelated research to fill a template.
© petekp · MIT · petekp/claude-code-setup
Similar skills
All skills for Agent setup: memory, context, orchestration →deep-research by other authors comes first.