Skip to content
Back to search

deep-research

Use when the user needs multi-source research with citation tracking, evidence persistence, and structured report generation. Triggers on "deep research", "comprehensive analysis", "research report", "compare X vs Y", "analyze trends", or "state of the art". Not for simple lookups, debugging, or questions answerable with 1-2 searches.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/199-biotechnologies-claude-deep-research-skill/badge.svg)](/listings/199-biotechnologies-claude-deep-research-skill)

HTML

<a href="/listings/199-biotechnologies-claude-deep-research-skill"><img src="/api/listings/199-biotechnologies-claude-deep-research-skill/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

# Clone into Claude Code skills directory
git clone https://github.com/199-biotechnologies/claude-deep-research-skill.git ~/.claude/skills/deep-research

No additional dependencies required for basic usage.

Optional: search-cli (multi-provider search)

For aggregated search across Brave, Serper, Exa, Jina, and Firecrawl:

brew tap 199-biotechnologies/tap && brew install search-cli
search config set keys.brave YOUR_KEY  # configure at least one provider

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

SKILL.md · Manifest · 108 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. ---
  2. name: deep-research
  3. description: Use when the user needs multi-source research with citation tracking, evidence persistence, and structured report generation. Triggers on "deep research", "comprehensive analysis", "research report", "compare X vs Y", "analyze trends", or "state of the art". Not for simple lookups, debugging, or questions answerable with 1-2 searches.
  4. ---
  5. # Deep Research
  6. ## Core Purpose
  7. Deliver citation-tracked research reports through a structured pipeline with evidence persistence, source identity management, claim-level verification, and progressive context management.
  8. **Autonomy Principle:** Operate independently. Infer assumptions from context. Only stop for critical errors or incomprehensible queries. Surface high-materiality assumptions explicitly in the Introduction and Methodology rather than silently defaulting.
  9. ---
  10. ## Decision Tree
  11. ```
  12. Request Analysis
  13. +-- Simple lookup? --> STOP: Use WebSearch
  14. +-- Debugging? --> STOP: Use standard tools
  15. +-- Complex analysis needed? --> CONTINUE
  16. Mode Selection
  17. +-- Initial exploration --> quick (3 phases, 2-5 min)
  18. +-- Standard research --> standard (6 phases, 5-10 min) [DEFAULT]
  19. +-- Critical decision --> deep (8 phases, 10-20 min)
  20. +-- Comprehensive review --> ultradeep (8+ phases, 20-45 min)
  21. ```
  22. **Default assumptions:** Technical query = technical audience. Comparison = balanced perspective. Trend = recent 1-2 years.
  23. ---
  24. ## Workflow Overview
  25. | Phase | Name | Quick | Std | Deep | Ultra |
  26. |-------|------|-------|-----|------|-------|
  27. | 1 | SCOPE | Y | Y | Y | Y |
  28. | 2 | PLAN | - | Y | Y | Y |
  29. | 3 | RETRIEVE | Y | Y | Y | Y |
  30. | 4 | TRIANGULATE | - | Y | Y | Y |
  31. | 4.5 | OUTLINE REFINEMENT | - | Y | Y | Y |
  32. | 5 | SYNTHESIZE | - | Y | Y | Y |
  33. | 6 | CRITIQUE | - | - | Y | Y |
  34. | 7 | REFINE | - | - | Y | Y |
  35. | 8 | PACKAGE | Y | Y | Y | Y |
  36. **Note:** Phases 3-5 operate as an evidence loop per section (retrieve → evidence store → refine outline → draft → verify claims → delta-retrieve if needed), not as strict sequential gates.
  37. ---
  38. ## Execution
  39. **On invocation, load relevant reference files:**
  40. 1. **Phase 1-7:** Load [methodology.md](./reference/methodology.md) for detailed phase instructions
  41. 2. **Phase 8 (Report):** Load [report-assembly.md](./reference/report-assembly.md) for progressive generation
  42. 3. **HTML/PDF output:** Load [html-generation.md](./reference/html-generation.md)
  43. 4. **Quality checks:** Load [quality-gates.md](./reference/quality-gates.md)
  44. 5. **Long reports (>18K words):** Load [continuation.md](./reference/continuation.md)
  45. **Templates:**
  46. - Report structure: [report_template.md](./templates/report_template.md)
  47. - HTML styling: [mckinsey_report_template.html](./templates/mckinsey_report_template.html)
  48. **Scripts:**
  49. - `python scripts/validate_report.py --report [path]`
  50. - `python scripts/verify_citations.py --report [path]`
  51. - `python scripts/md_to_html.py [markdown_path]`
  52. ---
  53. ## Output Contract
  54. **Required sections:**
  55. - Executive Summary (200-400 words)
  56. - Introduction (scope, methodology, assumptions)
  57. - Main Analysis (4-8 findings, 600-2,000 words each, cited)
  58. - Synthesis & Insights (patterns, implications)
  59. - Limitations & Caveats
  60. - Recommendations
  61. - Bibliography (COMPLETE - every citation, no placeholders)
  62. - Methodology Appendix
  63. **Output files (all to `~/Documents/[Topic]_Research_[YYYYMMDD]/`):**
  64. - Markdown (primary source of truth)
  65. - `sources.jsonl` — stable source registry with canonical IDs
  66. - `evidence.jsonl` — append-only evidence store with quotes and locators
  67. - `claims.jsonl` — atomic claim ledger with support status
  68. - `run_manifest.json` — query, mode, assumptions, provider config
  69. - HTML (McKinsey style, auto-opened)
  70. - PDF (professional print, auto-opened)
  71. **Quality standards:**
  72. - 10+ sources, 3+ per major claim (cluster-independent, not just count)
  73. - All factual claims cited immediately [N] with evidence backing in `evidence.jsonl`
  74. - Claim-support verification mandatory: no unsupported factual claims pass delivery
  75. - No placeholders, no fabricated citations
  76. - Prose-first (>=80%), bullets sparingly
  77. ---
  78. ## When to Use / NOT Use
  79. **Use:** Comprehensive analysis, technology comparisons, state-of-the-art reviews, multi-perspective investigation, market analysis.
  80. **Do NOT use:** Simple lookups, debugging, 1-2 search answers, quick time-sensitive queries.
README.md · README · 119 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

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

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.