Deep Research

Use this skill when the task requires thorough, multi-source research with citations — more depth than a quick web search.

Available Tools

Tool	Purpose	Depth
deep_research	Multi-step web search + AI analysis with citations	Comprehensive — generates full reports
semantic_content_search	Search site content via vector embeddings	Finds related content already on your WordPress site
search_content	Hybrid keyword + semantic search of site content	Use search_type: "hybrid" for best results across your content

WP-CLI Commands

Key mcp-ai WP-CLI commands for research workflows. Run via terminal:

Command	Use for
wp mcp-ai harness	Semantic search across stored content
wp mcp-ai health	Check search/tool connectivity
wp mcp-ai memory	Review stored research context

When to use this skill

Trigger when ANY of the following is true:

• The task asks to "research", "investigate", or "do a deep dive" on a topic.
• Building a brand or company profile for marketing purposes.
• Researching a market, industry trend, or competitive landscape.
• The task requires citations, sources, and substantiated findings.
• "Find out everything about X" — breadth + depth matters.
• Preparing a research report that will inform a campaign, content piece, or strategy.
• The task mentions "deep research", "comprehensive research", or "multi-source".

deep_research — Comprehensive Topic Research

Use for broad, multi-source investigation. Performs multi-step web searches, AI analysis, and generates a structured report with citations.

Parameters

topic          — The research topic or question (required)
depth          — "basic" | "standard" | "comprehensive" (default: "standard")
focus_areas    — Optional specific aspects: ["technical", "historical", "pricing", "competitors"]
include_sources — Include source citations (default: true)
run_mode       — "immediate" (synchronous) | "background" (WordPress cron)
include_site_content — Search WordPress + Vector Store alongside web (default: true)
memory_agent_id — Recall prior research from agent memory before starting
store_to_memory — Auto-store report in agent memory (requires memory_agent_id)

Depth Selection

Depth	When	Example
basic	Quick overview, fact-checking, simple questions	"What are the top fragrance brands in Sri Lanka?"
standard	Marketing research, brand profiles, campaign prep	"Research Conatural — history, products, target market, competitors"
comprehensive	Competitive analysis, market reports, strategic planning	"Full analysis of Sri Lanka's luxury fragrance e-commerce market"

Focus Areas

Guide the research by specifying aspects:

For a brand:    ["history", "product range", "pricing", "target audience", "competitors"]
For a market:   ["market size", "trends", "key players", "consumer behavior", "growth forecast"]
For a trend:    ["origins", "current state", "adoption", "future outlook", "Sri Lanka context"]
For a competitor: ["products", "pricing", "marketing channels", "USPs", "weaknesses"]

Run Modes

Mode	When	Time
immediate	Interactive tasks where you need results now	30s–2min
background	Long research, non-urgent, large scope	Minutes to hours

Use immediate for most Sophie tasks. Use background for comprehensive market reports or when researching 10+ entities.

Output

Returns:

• Report — structured findings with sections
• Sources — URLs with type labels (manufacturer, retailer, news, etc.)
• Product data (if applicable) — specs, pricing, images, attributes
• SEO keywords — extracted from research for content optimization

Example

deep_research {
  topic: "Conatural skincare brand — history, product range, target market,
          certifications, competitors in Sri Lanka"
  depth: "standard"
  focus_areas: ["brand history", "product categories", "certifications",
                "pricing strategy", "Sri Lanka market position"]
  include_sources: true
  run_mode: "immediate"
}

semantic_content_search — Site Knowledge Retrieval

Search your own WordPress content (posts, pages, products) by meaning, not just keywords. Uses vector embeddings for semantic matching.

Parameters

query          — Natural language search query
post_types     — Post types to search (default: ["post", "page"])
limit          — Max results (default: 10, max: 100)
threshold      — Minimum similarity score 0–1 (default: 0.7)
vector_store_id — Optional OpenAI vector store ID
include_meta   — Include post metadata (default: false)

When to use

• "Have we written about this before?" — check for existing content
• "What products are related to X?" — find semantically similar products
• "Find all content about Sri Lankan luxury market" — thematic search
• Before creating new content — avoid duplication

Example

semantic_content_search {
  query: "luxury fragrance gifting ideas for Sri Lankan customers"
  post_types: ["post", "product"]
  threshold: 0.75
  limit: 10
}

Research Workflow

For a typical research task, follow this sequence:

1. SITE CHECK (optional)
   semantic_content_search → "Do we already have content on this?"

2. DEEP RESEARCH
   deep_research {
     topic: "..."
     depth: "standard"
     include_sources: true
   }

3. EXTRACT INSIGHTS
   From the report, pull:
   ● Key facts (for content)
   ● Competitor intel (for positioning)
   ● Market gaps (for campaign angles)
   ● Sources (for citations and further reading)

4. ACT ON IT
   → Feed findings into campaign orchestration
   → Create a WordPress post via create_post_from_research
   → Save to Paper Store for future reference

Paper Store Integration (on request)

When explicitly asked to persist research:

paper_store_write {
  collection: "research-reports"
  id: "topic-slug"
  title: "Research: [Topic]"
  connection_id: "conn_8srcad8zylhe"
  tags: ["deep-research", "topic-tag", "brand-tag"]
  body: {
    topic: "..."
    report: "..."
    key_findings: [...]
    sources: [...]
    researched_at: "YYYY-MM-DD"
  }
}

Critical Rules

• Match depth to need — a brand overview doesn't need comprehensive. Save that for competitive analysis.
• Always include sources — citations make research actionable and credible.
• Use focus areas — they dramatically improve report quality by guiding the AI.
• Check site content first — semantic_content_search prevents duplicating work already done.
• Feed research into action — research without application is wasted. Every report should inform a decision, campaign, or piece of content.
• Store when asked — persist to Paper Store only when the user requests it.

Common Mistakes

WRONG:
  ● Using deep_research for "What's 2+2?" (use web_search for quick facts)
  ● Running comprehensive depth for a 2-minute task
  ● Not specifying focus_areas (generic report, misses what you need)
  ● Researching without a purpose — no link to campaign, content, or decision
  ● Saving every research to Paper Store automatically

RIGHT:
  ✅ deep_research for multi-source, cited, structured reports
  ✅ web_search for quick fact checks
  ✅ standard depth for most marketing research tasks
  ✅ Specific focus_areas tied to what you'll do with the results
  ✅ Research → insight → action chain
  ✅ Paper Store only when explicitly requested

What This Skill Does NOT Cover

• Quick fact-checking — use design-web-research for fast lookups via web_search or brave_web_search.
• Product-specific research — use design-product-research for structured SKU/catalog investigations via research_product.
• Content writing — use design-content-research to convert research into WordPress draft posts via create_post_from_research.
• Campaign planning — use design-campaign-orchestration to apply research insights to campaign themes and calendars.
• SEO implementation — use design-seo-content to apply research keywords to product/category pages.
• Image or visual research — use analyze_image or design-image-generation for visual reference gathering.

Cross-references

• Run design-web-research for quick web searches — use web_search or brave_web_search instead of deep_research.
• Run design-product-research for structured product-specific research — use research_product.
• Run design-content-research to convert research into WordPress drafts — use create_post_from_research.
• Run design-campaign-orchestration to apply research insights to campaign planning.
• Run design-seo-content to use research keywords in product and category pages.
• Use paper_store_* with connection_id to persist research reports to specific sites.