| name | deep-research |
| description | Conduct enterprise-grade research with multi-source synthesis, citation tracking, and verification. Use when user needs comprehensive analysis requiring 10+ sources, verified claims, or comparison of approaches. Triggers include "deep research", "comprehensive analysis", "research report", "compare X vs Y", or "analyze trends". Do NOT use for simple lookups, debugging, or questions answerable with 1-2 searches. |
| allowed-tools | WebSearch, Read, Write, Bash(python:*) |
Deep Research
Conduct comprehensive, citation-backed research through multi-phase pipeline.
When to Use
Use when:
- Comprehensive analysis (10+ sources needed)
- Comparing technologies/approaches/strategies
- State-of-the-art reviews
- Multi-perspective investigation
- Technical decisions
- Market/trend analysis
Do NOT use:
- Simple lookups (use WebSearch)
- Debugging (use standard tools)
- 1-2 search answers
Research Modes
| Mode | Time | Sources | Use Case |
|---|
| Quick | 2-5 min | 5-10 | Exploration, time-sensitive |
| Standard | 5-10 min | 15-30 | Most use cases [DEFAULT] |
| Deep | 10-20 min | 30-50 | Important decisions |
| UltraDeep | 20-45 min | 50-100 | Critical analysis |
Workflow
Phase 1: SCOPE
Define research boundaries:
- Core question
- Key sub-questions
- Required perspectives
- Success criteria
Phase 2: PLAN
Formulate search strategy:
- Decompose into search angles
- Identify source types needed
- Plan parallel searches
Phase 3: RETRIEVE
CRITICAL: Parallel execution
Launch 5-10 searches simultaneously:
[Single message with multiple parallel tool calls]
WebSearch #1: Core topic semantic
WebSearch #2: Technical keywords
WebSearch #3: Recent 2024-2025 filtered
WebSearch #4: Academic domains
WebSearch #5: Critical analysis
WebSearch #6: Industry trends
Phase 4: TRIANGULATE
Verify claims with 3+ sources:
- Cross-reference facts
- Score source credibility
- Flag contradictions
Phase 5: SYNTHESIZE
Generate novel insights:
- Identify patterns
- Draw connections
- Extract implications
Phase 6: CRITIQUE
Red-team analysis:
- Challenge assumptions
- Identify gaps
- Consider alternatives
Phase 7: REFINE
Address weaknesses:
- Fill gaps
- Strengthen claims
- Improve citations
Phase 8: PACKAGE
Generate comprehensive report.
Anti-Hallucination Protocol
CRITICAL - Every claim must cite sources:
- ✅ GOOD: "Smith et al. (2024) found a 23% reduction [1]."
- ❌ BAD: "Studies show significant improvement."
- ✅ GOOD: "According to FDA data... [2]"
- ❌ BAD: "Research suggests..."
When uncertain:
- Say "No sources found for X" rather than fabricating
- Mark inferences as "This suggests..." not "Research shows..."
Report Structure
# Research Report: [Topic]
## Executive Summary
[50-250 words, key findings only]
## Introduction
[Research question, scope, methodology]
## Main Findings
### Finding 1: [Title]
[300-500 words with citations [1], [2], [3]]
### Finding 2: [Title]
[300-500 words with citations]
...
## Synthesis & Insights
[Novel patterns and implications]
## Limitations & Caveats
[Gaps, assumptions, uncertainties]
## Recommendations
[Actionable next steps]
## Bibliography
[1] Author (Year). "Title". Publication. URL
[2] ...
Quality Gates
- Minimum 10+ sources
- 3+ sources per major claim
- Executive summary <250 words
- Full citations with URLs
- No placeholders (TBD, TODO)
- Clear facts vs. analysis distinction
Writing Standards
- Narrative-driven: Flowing prose, not bullet lists
- Precision: Exact numbers embedded in sentences
- Economy: No fluff, respect reader's time
- Clarity: Technical terms, avoid ambiguity
Example:
- Bad: "significantly improved" → Good: "reduced mortality 23% (p<0.01)"
- Bad: "• Market: $2.4B" → Good: "The market reached $2.4 billion in 2023 [1]."
