content-research

Installation

SKILL.md

Content Research — Orchestrator

Research — Pre-creation intelligence. Discovers what content to create by analyzing what's winning in the market right now.

Core Question: "What content should we create, based on what's actually working for competitors and resonating with the audience?"

Critical Gates — Read First

Research informs — it does not decide. Output is evidence and patterns, not a content calendar. The user decides what to act on.
Recency is everything. Content research older than 30 days is stale. Ad creative data older than 14 days should be flagged. Social trends older than 7 days may have already peaked.
Longevity = performance proxy. An ad running for 60+ days is almost certainly profitable. An ad that ran for 3 days and stopped probably failed. Use run duration as the primary performance signal when engagement metrics aren't available.
Platform-native, not cross-posted. Research what works ON each platform. A winning LinkedIn post structure is useless on TikTok. Always tag findings by platform.
Use live data, not training data. Do NOT rely on memorized knowledge for competitor content, ad creatives, or trending topics. Always use WebSearch, WebFetch, Exa, Tavily, or other research tools.

Philosophy

The best content teams don't create in a vacuum — they research before they write. This skill bridges the gap between strategic research (market-research, icp-research) and short-form content execution (content-short, copywriting). It answers: "What short-form is winning, what hooks stop the scroll, and what should we ship?"

For long-form research (blog, case study, byline, PR, newsletter), use content-research-long — it covers SERP analysis, keyword intent, expert/source discovery, and dwell-time audience language that this skill does not.

Evidence > Intuition > Imitation.

When to use this vs. `content-research-long`

If your target is...	Use
TikTok/Reels script, ad creative, social post (LinkedIn/X/IG/FB), thread, carousel, SMS, OOH, drip email	content-research (this skill)
Blog post, case study, byline, press release, newsletter, app store/G2 listing, in-depth Quora answer, evergreen guide	content-research-long

The two skills share upstream artifacts (product-context.md, icp-research.md) and feed two different downstream skills (content-short vs. content-long). Run both if a campaign spans both formats.

Inputs Required

Niche, industry, or product category (required)
Competitor names or domains (optional — will discover if not provided)
Target platforms/channels (optional — will research all if not specified)

Output

.agents/mkt/content-research.md

Chain Position

Previous: icp-research, market-research (optional but recommended) | Next: imc-plan, content-create, copywriting

Skill Deference

Need to know what SHORT-FORM content to create based on what's winning? → Use this skill.
Need to know what LONG-FORM content to create (blog, case study, byline, PR, newsletter)? → Use content-research-long instead — it covers SERP analysis, keyword intent, source discovery, dwell-time audience language.
Need to understand the overall MARKET and competitive dynamics? → Use market-research instead — it maps markets, not content.
Need to understand WHO the customer is? → Use icp-research instead — it builds personas, not content briefs.
Ready to WRITE short-form content? → Use content-short instead — it executes, this skill researches.
Need to audit EXISTING pages for SEO? → Use seo instead — it audits what's live, this researches what to create.

Re-run triggers: Before any content campaign, before entering a new channel, monthly for active content programs, when competitor creative strategy shifts noticeably.

Agent Manifest

7 agents across 2 layers:

Agent	Layer	Role	Input	Output
ad-intel-agent	L1 (parallel)	Researches competitor ads across platforms — hooks, offers, formats, run duration	brief + competitors + platforms	Ad Creative Intelligence section
listening-agent	L1 (parallel)	Mines communities, reviews, forums for authentic audience language and pain points	brief + niche + communities	Audience Listening section
trend-agent	L1 (parallel)	Identifies emerging topics and trending formats across platforms	brief + niche + platforms	Trending Topics section
competitive-content-agent	L1 (parallel)	Analyzes competitor organic content — top performers, gaps, formats, frequency	brief + competitors	Competitive Content Audit section
pattern-agent	L2 (sequential)	Synthesizes L1 outputs into winning patterns and content opportunities	merged L1 outputs	Patterns & Opportunities section
brief-agent	L2 (sequential)	Converts opportunities into actionable content briefs with platform specs	pattern-agent output + L1 context	Content Brief Recommendations section
critic-agent	L2 (sequential)	Validates data freshness, source quality, signal vs. noise, actionability	full merged artifact	PASS or FAIL with rewrite instructions

Routing Logic

Route A: Ad Creative Intel

Trigger: "What ads are competitors running?", "Research the Meta ad library for [niche]", "Swipe file for [category]", "What creatives are working?"

ad-intel-agent → pattern-agent → critic-agent

Skip listening, trend, and competitive-content agents. Focus on paid creative intelligence only.

Route B: Audience Listening

Trigger: "What is the audience saying?", "What questions do they ask?", "Reddit research for [topic]", "Find authentic language for [niche]"

listening-agent → pattern-agent → critic-agent

Focus on community mining and audience language extraction.

Route C: Trend Research

Trigger: "What's trending in [niche]?", "Emerging topics for [category]", "What should we post about?"

trend-agent ──────────┐
                      ├──→ pattern-agent → critic-agent
competitive-content-agent┘

Combines trend detection with competitive content analysis to find trending topics competitors haven't covered yet.

Route D: Competitive Content Audit

Trigger: "What content are competitors publishing?", "Content gap analysis", "What are they ranking for that we're not?"

competitive-content-agent → pattern-agent → critic-agent

Focus on competitor organic content — blog posts, videos, social, newsletters.

Route E: Full Research Brief

Trigger: "Full content research for [niche]", "Research before our next content sprint", "What should we create next?"

ad-intel-agent ───────────┐
listening-agent ──────────┤
trend-agent ──────────────┼──→ pattern-agent → brief-agent → critic-agent
competitive-content-agent ┘

All 4 L1 agents run in parallel. Full L2 sequence produces actionable content briefs.

Scope Calibration

Decision Context	Research Depth	Competitor Count	Platforms
Quick check (single channel)	1-2 agents, surface research	2-3 competitors	1-2 platforms
Campaign prep	3-4 agents, moderate depth	3-5 competitors	2-4 platforms
Full content sprint	All agents, comprehensive	5-8 competitors	All relevant platforms

Dispatch Protocol

Step 0: Context Gathering

Check for upstream artifacts in this priority order:

research/product-context.md — Product category, differentiator, target audience (created by icp-research)
research/icp-research.md — Full persona profiles, VoC data, audience habitats
.agents/mkt/imc-plan.md (optional bonus — not upstream) — If it already exists from a prior run, use channel strategy to focus research on relevant platforms. Content-research is typically run before imc-plan, so this will often be null.

If none of the required artifacts (1-2) exist: interview for the minimum viable context:

What's the product/niche? — Category, differentiator, and who it's for
Who are the competitors? — At least 2-3 names or domains. If unknown, the competitive-content-agent and ad-intel-agent will discover them.
Which platforms matter? — Where does the audience spend time? If unknown, research all major platforms.
What's the goal? — Brand awareness? Lead gen? Sales? This determines which content types to prioritize.

Step 1: Route Selection

Based on the user's request and available context, select Route A-E.

If the user's request is vague ("research content for us"), default to Route E (Full Research Brief).

Single-Agent Fallback

If the task is narrow enough for a single agent (e.g., "what are competitors running on Meta?"), dispatch only the relevant agent without the full pipeline. Skip pattern-agent, brief-agent, and critic. Return the single agent's output directly.

Layer 1 Dispatch — Parallel Research Agents

Build the pre-writing context from Step 0:

pre-writing = {
  niche: [product category / industry],
  competitors: [list of competitor names/domains],
  platforms: [list of target platforms],
  goal: [awareness / leads / sales / engagement],
  product_context: [from product-context.md or interview],
  icp_data: [from icp-research.md or null],
  existing_channels: [from imc-plan.md if it exists, or null — optional bonus context]
}

Dispatch applicable L1 agents in parallel:

Agent	Route A	Route B	Route C	Route D	Route E
ad-intel-agent	Yes	—	—	—	Yes
listening-agent	—	Yes	—	—	Yes
trend-agent	—	—	Yes	—	Yes
competitive-content-agent	—	—	Yes	Yes	Yes

Research Tool Priority (all L1 agents)

Exa MCP (mcp__exa__web_search_exa) — best for semantic content search, finding similar content
Tavily MCP (mcp__tavily-remote-mcp__tavily_search, tavily_extract) — structured web search with content extraction
WebSearch / WebFetch — always available as fallback
Platform-specific access:
- Meta Ad Library: facebook.com/ads/library (free, no API key required — WebFetch)
- TikTok Creative Center: ads.tiktok.com/business/creativecenter (browse top ads)
- LinkedIn Ad Library: linkedin.com/ad-library (search by company)
- Reddit: direct API access or Exa/Tavily search with site:reddit.com
- YouTube: YouTube Data API or Exa search with site:youtube.com

Research Checkpoint

After L1 agents return, present findings summary and gather feedback before L2:

"Here's what I found across [platforms]. Before I identify patterns and build briefs:"

Any competitors I missed? Names or domains to add.
Any platforms I should skip or focus on? Priority channels.
Anything surprising that I should dig deeper on?
Do you have internal performance data? Past content metrics, best-performing posts, email open rates — this dramatically improves pattern identification.

Incorporate feedback into the merged L1 output before dispatching L2.

Layer 2 Dispatch — Sequential Synthesis

Step 1: Pattern Agent

dispatch pattern-agent:
  upstream: [merged output from all L1 agents]

Receives all L1 outputs. Identifies cross-platform patterns, winning formulas, content gaps, and opportunity areas.

Step 2: Brief Agent (Route E only)

dispatch brief-agent:
  upstream: [pattern-agent output + merged L1 context]
  references: [references/content-brief-template.md]

Converts patterns into 3-5 actionable content briefs, each specifying: topic, angle, format, platform, hook approach, evidence, and priority.

Step 3: Critic Agent

dispatch critic-agent:
  upstream: [full merged artifact — all L1 + L2 outputs]

Evaluates against quality gate. Returns PASS or FAIL.

Critic Gate — Max 2 Cycles

cycle = 0
while cycle < 2:
  verdict = critic-agent.evaluate(artifact)
  if verdict == PASS:
    break
  else:
    for each failure:
      re-dispatch named agent with feedback
    merge fixes into artifact
    cycle += 1

if cycle == 2 and verdict == FAIL:
  deliver artifact with critic's remaining notes as "[REVIEWER NOTE]" annotations
  warn user: "Artifact delivered with quality notes — some items could not be resolved in 2 cycles."

On rewrite: Only re-dispatch agents the critic names. Do not re-run the entire pipeline.

Quality Gate

Before delivering, verify the merged artifact passes all checks:

Artifact Template

On re-run: rename existing artifact to content-research.v[N].md and create new with incremented version.

---
skill: content-research
version: 1
date: {{today}}
status: draft
---

# Content Research: [Niche / Category]

**Date:** [today]
**Skill:** content-research
**Route:** [A / B / C / D / E]
**Competitors analyzed:** [list]
**Platforms covered:** [list]

## Ad Creative Intelligence

### Active Campaigns Overview

| Competitor | Platform | Format | Hook Type | Offer/CTA | Est. Run Duration | Status |
|-----------|----------|--------|-----------|-----------|-------------------|--------|
| [Name] | [Meta/TikTok/LinkedIn/YouTube] | [Video/Image/Carousel] | [Question/Bold claim/Social proof/How-to/UGC] | [Offer text] | [Days active] | Active/Paused |

### Winning Creative Patterns

| Pattern | Frequency | Platforms | Example | Why It Works |
|---------|-----------|-----------|---------|-------------|
| [Hook type / format / angle] | [How many competitors use it] | [Where] | [Specific example] | [Evidence-based reasoning] |

### Long-Running Ads (60+ days — high-confidence performers)

| Competitor | Platform | Hook | Key Message | Run Duration | Takeaway |
|-----------|----------|------|-------------|-------------|----------|

### Hook Analysis

| Hook Type | Count | Best Example | Platform | Why It Works |
|-----------|-------|-------------|----------|-------------|

### Offer & CTA Patterns

| CTA Type | Competitors Using | Platform | Destination |
|----------|------------------|----------|-------------|

## Audience Listening

### Top Questions & Pain Points

| Question / Pain Point | Source | Engagement Signal | Frequency | Verbatim Quote |
|----------------------|--------|-------------------|-----------|---------------|
| [What the audience asks/feels] | [Subreddit / forum / review site + link] | [Upvotes / replies / stars] | [How often this appears] | "[Exact words used]" |

### Authentic Language Map

| Concept (Brand Language) | Audience Language | Source | Usage Context |
|-------------------------|-------------------|--------|--------------|
| [How we say it] | [How they say it] | [Where found] | [When they use this phrase] |

### Emotional Triggers

| Trigger | Evidence | Strength | Content Angle |
|---------|----------|----------|--------------|

### Community Landscape

| Community | Platform | Size | Activity | Relevance | Key Themes |
|-----------|----------|------|----------|-----------|-----------|

## Trending Topics

### Cross-Platform Trends (appearing on 2+ platforms)

| Topic | Platforms | Velocity | Peak Window | Our Angle |
|-------|----------|----------|-------------|-----------|
| [Topic] | [Reddit + TikTok + YouTube...] | [Rising/Peaking/Declining] | [Estimated days remaining] | [How we'd cover this differently] |

### Platform-Specific Signals (single platform)

| Platform | Signal | Evidence | Risk Level |
|----------|--------|----------|------------|
| [Platform] | [Trending topic/format/sound] | [URL or data point] | [Low: established / Med: emerging / High: may peak soon] |

### Emerging Formats

| Format | Platform | Adoption Stage | Example | Opportunity |
|--------|----------|---------------|---------|------------|

### Trending Keywords & Hashtags

| Keyword/Hashtag | Platform | Volume/Growth | Relevance |
|----------------|----------|--------------|-----------|

## Competitive Content Audit

### Competitor Content Mix

| Competitor | Platform | Content Types | Frequency | Audience Size | Top Performer | Engagement |
|-----------|----------|--------------|-----------|--------------|---------------|-----------|

### Content Gap Matrix

| Topic / Angle | Us | Competitor A | Competitor B | Competitor C | Demand Signal |
|--------------|-----|-------------|-------------|-------------|--------------|
| [Topic] | [Covered/Missing] | [Covered/Missing] | [Covered/Missing] | [Covered/Missing] | [Search vol / community interest] |

### Top-Performing Competitor Content (last 90 days)

| Competitor | Content | Platform | Format | Engagement | Why It Worked |
|-----------|---------|----------|--------|-----------|---------------|

### Publishing Patterns

| Competitor | Frequency | Best Day/Time | Formats Used | Content Pillars |
|-----------|-----------|---------------|-------------|----------------|

### Content Quality Assessment

| Competitor | Strengths | Weaknesses | Unique Approach |
|-----------|-----------|-----------|----------------|

## Patterns & Opportunities

### Winning Patterns (cross-cutting)

| # | Pattern | Evidence | Platforms | Confidence |
|---|---------|----------|-----------|-----------|
| 1 | [Pattern description] | [N competitors use this, M community signals] | [Where] | H / M / L |

### Content Opportunities (ranked by evidence strength)

| # | Opportunity | Type | Evidence | Platform | Priority |
|---|------------|------|----------|----------|----------|
| 1 | [What to create] | [Gap / Trend / Audience need / Format] | [Supporting data] | [Where to publish] | High / Med / Low |

### Anti-Patterns to Avoid

| Pattern | Why It Fails | Evidence |
|---------|-------------|----------|

### Signal Strength Map

| Signal Source | Data Point | Supports Opportunity # |
|--------------|-----------|----------------------|

## Content Briefs (Route E only)

### Brief 1: [Title]

- **Topic:** [Specific topic]
- **Angle:** [Unique angle — what makes ours different from what exists]
- **Format:** [Blog / Video / Carousel / Thread / Ad / etc.]
- **Platform:** [Primary platform + secondary distribution]
- **Hook approach:** [Question / Bold claim / Social proof / How-to / Story / Data] — based on [winning pattern #N]
- **Key message:** [Core idea in one sentence]
- **Audience language to use:** [3-5 specific phrases from listening research]
- **Audience language to avoid:** [Brand jargon the audience doesn't use]
- **Evidence:** [Why this will work — cite specific research findings]
- **Competitive positioning:** [What competitors do/don't cover on this topic]
- **Priority:** [High / Med / Low] — [Justification]
- **Suggested execution:** `content-create` Route [A/B/D] → Channel: [specific channel]
- **Tracking:** [Suggested UTM structure or attribution approach]

### Brief 2: [Title]
[Same structure]

### Brief 3: [Title]
[Same structure]

### Brief Prioritization Summary

| # | Brief | Priority | Evidence Signals | Est. Effort | Impact |
|---|-------|----------|-----------------|-------------|--------|

## Research Limitations

| Aspect | Limitation | Implication |
|--------|-----------|-------------|
| [Data source] | [What couldn't be accessed] | [How this affects findings] |

## Next Step

Run `content-create` with the content briefs above, or `imc-plan` to incorporate these findings into a full channel strategy. Run `attribution` to set up tracking before publishing.

Worked Example

User: "Research content opportunities for our AI code review tool before our next content sprint."

Route E (Full Research Brief) with 5 competitors (CodeRabbit, Sourcery, Codacy, Qodo, Ellipsis). Product context: AI code review, multi-repo awareness differentiator, targeting engineering teams.

L1 parallel results: ad-intel found CodeRabbit's 78-day carousel ("5 things your code reviewer misses") as high-confidence winner. Listening mined "review tax" and "PR bottleneck" as authentic audience phrases from r/ExperiencedDevs. Trends detected "AI dev tools backlash" cross-platform and LinkedIn "hot take" format 3x outperformance. Competitive audit found multi-repo content gap and zero user-led (vs. founder-led) video testimonials.

Research checkpoint: User confirmed multi-repo as sweet spot, shared internal data (top blog post: "reducing PR cycle time", 8K views).

L2 synthesis: pattern-agent identified 5 cross-cutting patterns. brief-agent produced 3 briefs:

"The Review Tax Calculator" — interactive tool + blog, using "review tax" audience language
"Hot Take: Most AI Code Review Tools Miss 80% of What Matters" — contrarian LinkedIn post on multi-repo gap
"Engineering Lead Video Testimonial Series" — 60-sec clips, LinkedIn + YouTube Shorts

Critic: PASS. "Strong evidence-to-brief connection."

Anti-Patterns

Researching without a niche — "Research trending content" with no product category produces a BuzzFeed listicle, not actionable intelligence. Always anchor research to a specific niche and audience. INSTEAD: Require niche/category before dispatch. If the user doesn't provide one, pull from product-context.md or interview.

Copying competitor content — Research identifies what's working so you can create something BETTER with a unique angle, not so you can clone it. INSTEAD: pattern-agent identifies the underlying principle (why it works), not a template to copy. brief-agent always specifies a unique angle.

Confusing signals with trends — A single viral post is a signal. A pattern across 2+ platforms is a trend. Treating signals as trends leads to chasing noise. INSTEAD: trend-agent requires 2+ platform appearances before classifying anything as a "trend." Single-platform findings are labeled "signals."

Ignoring ad longevity — Engagement metrics (likes, shares) are vanity. Run duration is the real performance proxy — nobody keeps running unprofitable ads. INSTEAD: ad-intel-agent prioritizes run duration over engagement metrics. 60+ day ads get a dedicated "high-confidence performers" section.

Stale research treated as current — Content landscapes change weekly. A 60-day-old competitive analysis may be completely wrong. INSTEAD: Every finding includes a date. The 30-day staleness gate is enforced by the critic.

Platform-blind recommendations — "Post this on social media" is not actionable. Each platform has different winning formats, audiences, and algorithms. INSTEAD: Every finding and every brief is tagged by specific platform. No cross-platform generalizations without evidence.

Research without action — A beautiful research document that doesn't lead to content creation is waste. INSTEAD: Route E always produces content briefs. Every brief includes a suggested content-create route.

Required Artifacts

None — this is an entry point. Works best with upstream artifacts but functions without them.

Optional Artifacts

Artifact	Source	Benefit
`product-context.md`	icp-research	Product context for better competitor identification and angle development
`icp-research.md`	icp-research	Persona data for audience-language matching and community identification
`imc-plan.md`	imc-plan	Channel strategy for focused platform research
`market-research.md`	market-research	Competitor list and positioning for competitive content analysis

References

references/content-brief-template.md — Structured content brief format with platform-specific fields
references/ad-library-guide.md — How to access and interpret ad libraries across platforms
references/platform-research-patterns.md — Search queries and data extraction patterns per platform

Related skills

More from hungv47/research-skills

Installs

Repository

hungv47/research-skills

GitHub Stars

First Seen

Apr 13, 2026