1# Deep Research Skill for Claude Code
2 
3Enterprise-grade research engine for Claude Code. Produces citation-backed reports with source credibility scoring, multi-provider search, and automated validation.
4 
5## Installation
6 
7```bash
8# Clone into Claude Code skills directory
9git clone https://github.com/199-biotechnologies/claude-deep-research-skill.git ~/.claude/skills/deep-research
10```
11 
12No additional dependencies required for basic usage.
13 
14### Optional: search-cli (multi-provider search)
15 
16For aggregated search across Brave, Serper, Exa, Jina, and Firecrawl:
17 
18```bash
19brew tap 199-biotechnologies/tap && brew install search-cli
20search config set keys.brave YOUR_KEY  # configure at least one provider
21```
22 
23## Usage
24 
25```
26deep research on the current state of quantum computing
27```
28 
29```
30deep research in ultradeep mode: compare PostgreSQL vs Supabase for our stack
31```
32 
33## Research Modes
34 
35| Mode | Phases | Duration | Best For |
36|------|--------|----------|----------|
37| Quick | 3 | 2-5 min | Initial exploration |
38| Standard | 6 | 5-10 min | Most research questions |
39| Deep | 8 | 10-20 min | Complex topics, critical decisions |
40| UltraDeep | 8+ | 20-45 min | Comprehensive reports, maximum rigor |
41 
42## Pipeline
43 
44Scope → Plan → **Retrieve** (parallel search + agents) → Triangulate → Outline Refinement → Synthesize → Critique (with loop-back) → Refine → Package
45 
46Key features:
47- **Step 0**: Retrieves current date before searches (prevents stale training-data year assumptions)
48- **Parallel retrieval**: 5-10 concurrent searches + 2-3 focused sub-agents returning structured evidence objects
49- **First Finish Search**: Adaptive quality thresholds by mode
50- **Critique loop-back**: Phase 6 can return to Phase 3 with delta-queries if critical gaps found
51- **Multi-persona red teaming**: Skeptical Practitioner, Adversarial Reviewer, Implementation Engineer (Deep/UltraDeep)
52- **Disk-persisted citations**: `sources.json` survives context compaction and continuation agents
53 
54## Output
55 
56Reports saved to `~/Documents/[Topic]_Research_[Date]/`:
57- Markdown (primary source of truth)
58- HTML (McKinsey-style, auto-opened in browser)
59- PDF (professional print via WeasyPrint)
60 
61Reports >18K words auto-continue via recursive agent spawning with context preservation.
62 
63## Quality Standards
64 
65- 10+ sources, 3+ per major claim
66- Executive summary 200-400 words
67- Findings 600-2,000 words each, prose-first (>=80%)
68- Full bibliography with URLs, no placeholders
69- Automated validation: `validate_report.py` (9 checks) + `verify_citations.py` (DOI/URL/hallucination detection)
70- Validation loop: validate → fix → retry (max 3 cycles)
71 
72## Search Tools
73 
74| Tool | Priority | Setup |
75|------|----------|-------|
76| search-cli | **Primary** — all searches go here first | `brew install search-cli` + API keys |
77| WebSearch | Fallback — if search-cli fails or rate-limited | None (built-in) |
78| Exa MCP | Optional — semantic/neural search alongside search-cli | MCP config |
79 
80## Architecture
81 
82```
83deep-research/
84├── SKILL.md                          # Skill entry point (lean, ~100 lines)
85├── reference/
86│   ├── methodology.md                # 8-phase pipeline details
87│   ├── report-assembly.md            # Progressive generation strategy
88│   ├── quality-gates.md              # Validation standards
89│   ├── html-generation.md            # McKinsey HTML conversion
90│   ├── continuation.md               # Auto-continuation protocol
91│   └── weasyprint_guidelines.md      # PDF generation
92├── templates/
93│   ├── report_template.md            # Report structure template
94│   └── mckinsey_report_template.html # HTML report template
95├── scripts/
96│   ├── validate_report.py            # 9-check structure validator
97│   ├── verify_citations.py           # DOI/URL/hallucination checker
98│   ├── source_evaluator.py           # Source credibility scoring
99│   ├── citation_manager.py           # Citation tracking
100│   ├── md_to_html.py                 # Markdown to HTML converter
101│   ├── verify_html.py                # HTML verification
102│   └── research_engine.py            # Core orchestration engine
103└── tests/
104    └── fixtures/                     # Test report fixtures
105```
106 
107## Version History
108 
109| Version | Date | Changes |
110|---------|------|---------|
111| 2.3.1 | 2026-03-19 | Template/validator harmonization, structured evidence, critique loop-back, multi-persona red teaming |
112| 2.3 | 2026-03-19 | Contract harmonization, search-cli integration, dynamic year detection, disk-persisted citations, validation loops |
113| 2.2 | 2025-11-05 | Auto-continuation system for unlimited length |
114| 2.1 | 2025-11-05 | Progressive file assembly |
115| 1.0 | 2025-11-04 | Initial release |
116 
117## License
118 
119MIT - modify as needed for your workflow.
120