| name | deep-research |
| description | Conducts enterprise-grade research with multi-source synthesis, citation tracking, and verification. Produces citation-backed reports through a structured pipeline. Uses FREE providers by default (WebSearch, SearXNG). 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.
|
Deep Research
Agent-orchestrated research pipeline. No Python dependencies required.
Core Purpose
Deliver citation-backed, verified research reports through a structured 8-phase pipeline using the agent's native capabilities.
Key Principle: The agent IS the research engine. Follow methodology rules, use LLM reasoning for evaluation, and leverage built-in tools for search and extraction.
Provider Priority
FREE-FIRST approach:
| Priority | Provider | Cost | Availability |
|---|
| 1 | WebSearch | FREE | Always (built-in) |
| 2 | SearXNG | FREE | Docker (optional) |
| 3 | Exa MCP | FREE tier | MCP config (optional) |
| 4 | search-cli | PAID | API keys (fallback only) |
Always use free options first. See providers.md for details.
Detection:
- WebSearch: Always available
- SearXNG:
curl -s http://localhost:8000/health
- Exa MCP: Check for
mcp__Exa__exa_search tool
Decision Tree
Request Analysis
+-- Simple lookup? --> STOP: Use WebSearch
+-- Debugging? --> STOP: Use standard tools
+-- Complex analysis needed? --> CONTINUE
Mode Selection
+-- Initial exploration --> quick (3 phases, 2-5 min)
+-- Standard research --> standard (6 phases, 5-10 min) [DEFAULT]
+-- Critical decision --> deep (8 phases, 10-20 min)
+-- Comprehensive review --> ultradeep (8+ phases, 20-45 min)
Default assumptions: Technical query = technical audience. Comparison = balanced perspective. Trend = recent 1-2 years.
Workflow Overview
| Phase | Name | Quick | Standard | Deep | UltraDeep |
|---|
| 1 | SCOPE | Y | Y | Y | Y |
| 2 | PLAN | - | Y | Y | Y |
| 3 | RETRIEVE | Y | Y | Y | Y |
| 4 | TRIANGULATE | - | Y | Y | Y |
| 4.5 | OUTLINE REFINEMENT | - | Y | Y | Y |
| 5 | SYNTHESIZE | - | Y | Y | Y |
| 6 | CRITIQUE | - | - | Y | Y |
| 7 | REFINE | - | - | Y | Y |
| 8 | PACKAGE | Y | Y | Y | Y |
Execution
On invocation, load ALL reference files:
- methodology.md - Complete 8-phase pipeline instructions
- providers.md - Search provider options and detection
- browser-scraping.md - Browser fallback extraction
- quality-gates.md - Validation checklists
- report-assembly.md - Report generation strategy
- html-generation.md - HTML report styling
- continuation.md - Long report handling
Templates:
No Python Scripts: All orchestration happens through agent following methodology rules.
Output Contract
Required sections:
- Executive Summary (200-400 words)
- Introduction (scope, methodology, assumptions)
- Main Analysis (4-8 findings, 600-2,000 words each, cited)
- Synthesis & Insights (patterns, implications)
- Limitations & Caveats
- Recommendations
- Bibliography (COMPLETE - every citation, no placeholders)
- Methodology Appendix
Output location: Use the active project's docs/research/ by default, or a location explicitly designated by the user. Never mix unrelated projects in shared research storage; if no project or approved location is clear, ask before writing.
Output files (all to <research_root>/[Topic]_Research_[YYYYMMDD]/):
- Markdown (primary source of truth)
- HTML (McKinsey style, optional)
Quality standards:
- 10+ sources, 3+ per major claim
- All claims cited immediately [N]
- No placeholders, no fabricated citations
- Prose-first (>=80%), bullets sparingly
Invocation Triggers
Use this skill when the user says:
- "deep research on [topic]"
- "comprehensive analysis of [topic]"
- "research report on [topic]"
- "compare [X] vs [Y]"
- "analyze trends in [topic]"
- "state of the art in [topic]"
- "investigate [topic] thoroughly"
- "what does the literature say about [topic]"
Do NOT trigger for:
- Simple lookups (use WebSearch directly)
- Debugging (use standard tools)
- Questions answerable with 1-2 searches
- Quick time-sensitive queries
Agent Responsibilities
The agent (not scripts) handles:
- Orchestration: Execute phases by reading methodology.md
- Source evaluation: Use LLM reasoning to score credibility
- Citation tracking: Maintain source provenance in context
- Validation: Check against quality-gates.md checklists
- Report generation: Create output following templates
No Dependencies
This skill requires:
- No Python runtime
- No pip install
- No external dependencies
All orchestration is handled by the agent following the methodology rules.