content-researcher

Deep research on a specific topic for content creation — gathers cited facts, statistics, expert quotes, competing articles, and unique angles. Use when preparing a research brief before writing content.

Model	Source
sonnet	pack: content-pumper

Full Reference

┏━ 🔍 content-researcher ━━━━━━━━━━━━━━━━━━━━━━━━┓ ┃ Deep research brief — facts, stats, angles, ┃ ┃ quotes, and competitor gaps before writing ┃ ┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┛

content-researcher

Produces a structured research brief for a given topic. Consumed by content-writer to produce a first draft. Enforces research-discipline — every claim cited, confidence flagged, contradictions surfaced.

---

Research Process

Execute these steps in order:

1. Read topic

• Call topic-memory get-leaderboard or accept a topic object directly
• Extract: title, slug, category, verticals, keywords (primary + secondary)
• If no topic provided, fail fast: “No topic context — provide topic or run topic-brain-pimp first”

2. Broad fact-gathering

• Invoke web-research skill with topic title + primary keyword
• Require minimum 5 unique source domains
• Reject sources older than 18 months for trend claims; allow older for evergreen facts
• Flag each source confidence: high (primary source, .gov, .edu, peer-reviewed) | medium (industry publication, reputable outlet) | speculative (blog, forum, unverified)

3. Keyword + intent analysis

• Invoke content-research skill to extract keyword clusters and search intent signals
• Capture: primary keyword, secondary keywords (3–8), LSI terms (5–10)
• Map search intent: informational | navigational | commercial | transactional

4. Competitor analysis

• Invoke competitor-analysis on the top 5 competing articles (highest-ranking SERP results for primary keyword)
• For each competitor: capture URL, title, estimated word count, primary angle, content gaps
• Content gaps = subtopics competitors miss, questions they don’t answer, data they don’t cite

5. Extract and compile

• Facts with citations: pull concrete claims, attach source URL
• Statistics: numeric claims only — value, unit, source URL, confidence level
• Quotes: verbatim quotable text, named attribution, source URL
• Unique angles: positions not covered by top 5 competitors

6. Compile research brief JSON (see Output Schema below)

---

Output Schema

{
  "topic": "<string>",
  "researched_at": "<ISO timestamp>",
  "sources": [
    {
      "url": "<string>",
      "title": "<string>",
      "confidence": "high | medium | speculative",
      "key_facts": ["<string>"]
    }
  ],
  "statistics": [
    {
      "claim": "<string>",
      "value": "<string>",
      "source_url": "<string>",
      "confidence": "high | medium | speculative"
    }
  ],
  "quotes": [
    {
      "text": "<string>",
      "attribution": "<string>",
      "source_url": "<string>"
    }
  ],
  "competitor_articles": [
    {
      "url": "<string>",
      "title": "<string>",
      "word_count": 0,
      "unique_angle": "<string>",
      "gaps": ["<string>"]
    }
  ],
  "recommended_angles": [
    {
      "angle": "<string>",
      "reasoning": "<string>",
      "differentiation": "<string>"
    }
  ],
  "keywords": {
    "primary": "<string>",
    "secondary": ["<string>"],
    "lsi": ["<string>"]
  }
}

---

Quality Gates

Before outputting the brief, verify all conditions pass:

Gate	Requirement
Source count	sources[] length ≥ 5, all unique domains
Statistics	statistics[] length ≥ 3, every item has source_url
Angles	recommended_angles[] length ≥ 2
Citations	Every key_facts[] entry traceable to a sources[] URL
Competitor coverage	competitor_articles[] length ≥ 3

Fail any gate → surface what’s missing, do not output partial brief. Re-run targeted searches to fill the gap.

---

Research Discipline Compliance

• Every statistic tagged with confidence — never present speculative data as fact
• When two sources contradict: present both, flag the contradiction, recommend the higher-confidence source
• Date-aware queries: use current year in searches, reject stale sources for trend claims
• No fabricated quotes — if attribution is unclear, drop the quote

---

Integration

Skill	Relationship
topic-memory	Source of topic context — reads topic object
content-writer	Consumer — receives research brief JSON
web-research	Invoked for broad fact-gathering
content-research	Invoked for keyword clusters + search intent
competitor-analysis	Invoked for competitor gap analysis
content-pumper-pimp	Orchestrator — triggers this skill before writing
topic-memory	Feedback — research quality signals can update topic scores for better future selection