deep-research

name	deep-research
version	2.0
description	Deep multi-source research synthesis — web, academic, patent, and internal KB cross-referencing
trigger	deep research, research synthesis, literature review, patent search, competitive analysis, intelligence gathering

Deep Research Skill v2.0

Purpose

Conduct thorough multi-source research across web, academic databases, patent systems, and internal Trinity KB, then synthesize findings into actionable intelligence.

Capabilities

• Multi-source parallel search (Tavily, arXiv, PubMed, Semantic Scholar, Wikipedia, OEIS)
• Patent landscape analysis
• Competitive intelligence gathering
• Academic literature review and citation mapping
• Internal KB cross-referencing for existing knowledge
• Executive summary generation with confidence scores

Research Protocol

1. Scope: Define research question, constraints, and target depth
2. Search: Fan out to 3+ sources in parallel
3. Filter: Rank by relevance, recency, and authority
4. Synthesize: Cross-reference findings, identify patterns and contradictions
5. Report: Structured output with citations, confidence levels, and knowledge gaps

Agents

• Researcher (primary): Deep web and academic search
• Oracle: Internal KB and data intelligence
• Prophet: Strategic pattern recognition

Tools

• deep_web_search (Tavily), research_topic
• patent_search, arxiv_search, pubmed_search, semantic_scholar_search
• wikipedia_lookup, wolfram_alpha_query
• search_vault (internal KB)

Quality Gates

• Minimum 3 independent sources for any factual claim
• All citations must include URL/DOI
• Confidence scoring: HIGH (3+ corroborating sources), MEDIUM (2), LOW (1)
• Research reports must include "Knowledge Gaps" section

Protocol

1. Assess — Understand the specific requirement and context
2. Plan — Determine the approach based on the guidance above
3. Execute — Apply the deep research methodology
4. Verify — Validate the output against expected standards
5. Report — Document results and any issues encountered

Integration

This skill is available via:

• invoke_skill("deep-research") in the agent kernel
• POST /api/skills/invoke with skillId: "deep-research"
• Semantic search via GET /api/skills/search?q=deep+research