deep-research

Generates citation-backed research reports through an 8-phase multi-source verification pipeline.

Updated Jun 10, 2026
One-click install
npx skills add https://github.com/eylulsenakumral/auto-company-clean --skill deep-research-eylulsenakumral
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: deep-research
Source: https://github.com/eylulsenakumral/auto-company-clean/tree/main/.claude/skills/deep-research
Command: npx skills add https://github.com/eylulsenakumral/auto-company-clean --skill deep-research-eylulsenakumral

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve? Producing trustworthy, comprehensive research reports requires gathering many sources, verifying claims across them, tracking citations, and structuring findings — a slow, error-prone manual process that often yields uncited or fabricated claims. ## Core Features & Use Cases - 8-Phase Research Pipeline: Scope, Plan, Retrieve, Triangulate, Synthesize, Critique, Refine, and Package phases with parallel web searches and sub-agent execution. - Source Credibility & Validation: Scores sources 0-100, enforces 3+ source triangulation per claim, and runs automated 8-check report validation plus citation verification. - Four Depth Modes: Quick (2-5 min), Standard (5-10 min), Deep (10-20 min), and UltraDeep (20-45 min) with progressive file assembly for reports up to 20,000+ words. - Multi-Format Output: Generates Markdown, McKinsey-style HTML, and PDF reports saved to an organized folder in ~/Documents. - Use Case: Ask for a comprehensive analysis comparing two technologies, and receive a fully cited report with executive summary, findings, limitations, recommendations, and complete bibliography. ## Quick Start Use deep research to compare PostgreSQL and Supabase for my project and generate a cited report.

Frequently Asked Questions about deep-research

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a deep research report in Claude Code?▼

Invoke the skill with a natural language request like "Use deep research to analyze AI agent frameworks in 2025". The skill auto-selects a mode, runs parallel web searches, verifies claims across sources, and saves Markdown, HTML, and PDF reports to a folder in ~/Documents.

What research modes does the deep research skill support?▼

Four modes: Quick (3 phases, 2-5 minutes), Standard (6 phases, 5-10 minutes, default), Deep (8 phases, 10-20 minutes), and UltraDeep (8+ phases, 20-45 minutes). Mode is selected automatically from query signals or can be specified explicitly.

How does the skill prevent fabricated citations?▼

Every factual claim must include an immediate [N] citation, and a verify_citations.py script checks DOI resolution and title/year matching to flag suspicious entries. Reports failing citation checks must be corrected before delivery.

Does the deep research skill require external Python packages?▼

No. All scripts (research_engine, validate_report, source_evaluator, citation_manager) use only the Python standard library, so no pip installation is needed. It requires Python 3 and access to the WebSearch tool.

When should I not use the deep research skill?▼

Avoid it for simple lookups, debugging tasks, or questions answerable with 1-2 searches — use a direct web search instead. It is designed for complex analysis needing 10+ sources and verified claims.

What happens if report validation fails?▼

The validator runs 8 automated checks covering summary length, required sections, citation formatting, bibliography completeness, and source count. On failure it attempts up to 2 fixes, then stops and reports the issues to the user.