Deep Research Skill for Claude Code

Enterprise-grade research engine for Claude Code. Produces citation-backed reports with source credibility scoring, multi-provider search, and automated validation.

Use Cases

• Research tasks ranging from quick exploration to comprehensive deep reports
• Studies requiring citation support and source credibility assessment
• Complex topics, critical decisions, and comprehensive reports

Core Capabilities

• Four research modes: Quick (3 phases, 2-5 min), Standard (6 phases, 5-10 min), Deep (8 phases, 10-20 min), UltraDeep (8+ phases, 20-45 min)
• 8-phase pipeline: Scope → Plan → Retrieve (parallel search + sub-agents) → Triangulate → Outline Refinement → Synthesize → Critique (with loop-back) → Refine → Package
• Parallel retrieval: 5-10 concurrent searches + 2-3 focused sub-agents returning structured evidence objects
• Adaptive quality thresholds (First Finish Search)
• Critique loop-back: Phase 6 can return to Phase 3 with delta-queries if critical gaps found
• Multi-persona red teaming (Deep/UltraDeep): Skeptical Practitioner, Adversarial Reviewer, Implementation Engineer
• Disk-persisted citations: sources.json survives context compaction and continuation agents
• Auto-continuation: reports >18,000 words auto-continue via recursive agent spawning with context preservation
• Quality standards: 10+ sources, 3+ per major claim; executive summary 200-400 words; each finding 600-2,000 words, prose-first (≥80%); full bibliography with URLs, no placeholders
• Automated validation: validate_report.py (9 checks) + verify_citations.py (DOI/URL/hallucination detection); validation loop: validate → fix → retry (max 3 cycles)

How It Works

1. You provide a research topic (as a text command)
2. The Skill executes:
   • Retrieves current date (prevents stale training-data year assumptions)
   • Runs the research pipeline according to the selected mode: retrieval, triangulation, synthesis, critique, and refinement
   • Generates reports in Markdown, HTML, and PDF formats
   • Automatically validates report quality and loops back to fix issues if needed
3. You receive the complete report saved in ~/Documents/[Topic]_Research_[Date]/

What You Need to Provide

• Research topic (text input)
• Optional: Install search-cli and configure at least one provider API key (Brave, Serper, Exa, Jina, Firecrawl) to enable multi-provider search
• Optional: Configure Exa MCP for semantic/neural search

What You Will Get

• Complete Markdown report (primary source of truth)
• HTML report (McKinsey-style, auto-opened in browser)
• PDF report (professional print via WeasyPrint)
• Report includes: executive summary, findings, full bibliography with URLs, source credibility scores
• Auto-validated report

Usage Examples

deep research on the current state of quantum computing

deep research in ultradeep mode: compare PostgreSQL vs Supabase for our stack

Mode Selection

Notes

• This Skill requires Claude Code environment.
• Basic usage has no additional dependencies. Optionally, install search-cli to enable multi-provider search (Brave, Serper, Exa, Jina, Firecrawl) with corresponding API keys.
• PDF output depends on WeasyPrint; ensure it is available on your system.
• Reports exceeding 18,000 words auto-continue via recursive agent spawning, which may increase processing time.
• The automated validation loop retries up to 3 cycles to ensure quality.