Skills/Competitive Intelligence/Competitor Research

Competitor Research

This skill should be used when the user wants to research competitors' Meta ads, analyze their creative strategies, and generate actionable intelligence. Triggers on requests like "research competitors", "competitor analysis", "analyze competitor ads", "competitive intelligence", or when the user provides a CSV of competitor names, page IDs, or keywords for ad research.

Competitor Research

Deep competitive intelligence from Meta Ads Library with visual creative analysis and actionable creative briefs.

Go beyond screenshots. This skill scrapes competitor ads from Meta Ads Library, downloads static image creatives, visually analyzes each one using Claude's multimodal vision, generates a deep competitive research report, and produces creative briefs with 15 hook variations per competitor — all in one run.

What it does

Recon Agent — Scrapes Meta Ads Library for each competitor via Glued MCP. Filters to IMAGE ads only, downloads each creative to a local images/ folder, and writes structured raw_data.json with image paths.

Analyst Agent — Reads every downloaded image using Claude's vision capabilities. Performs quantitative analysis (ad volume, runtime winners, platform distribution, launch cadence) plus visual creative analysis (composition, color palette, text overlays, product presentation, brand elements). Writes competitive_research.md with a dedicated Visual Creative Analysis section.

Strategist Agent — Turns research into action. Generates 3-5 creative briefs targeting specific gaps (Counter-Position, Gap Exploit, Winner Remix, Format Arbitrage, Hook Upgrade). Produces 15 hooks per competitor across all hook types. Adapts to brand voice if available.

Dashboard — Generates an interactive HTML dashboard with Chart.js showing ad volume, runtime analysis, platform distribution, top hooks, and CTA heatmap.

Output

output/<YYYYMMDD_HHmmss>_competitor_research/
├── images/                    # Downloaded static ad images

│   ├── AG1_1.jpg

│   ├── Huel_1.jpg

│   └── ...

├── competitive_research.md    # Deep analysis with visual creative breakdown

├── creative_briefs.md         # 3-5 actionable briefs (in brand voice if available)

├── hooks_matrix.csv           # 15 hooks x N competitors, structured

├── hooks.md                   # Human-readable hooks with rationale

├── dashboard.html             # Interactive Chart.js dashboard

└── raw_data.json              # IMAGE ads only, with image_path fields

What the analysis actually looks like

The visual creative analysis covers each competitor's top 10 ads:

Composition — Layout style, focal point, visual hierarchy
Color palette — Dominant colors, contrast, mood
Text overlays — Headline copy, font style, text-to-image ratio, placement
Product presentation — Hero shot, lifestyle, before/after, flat lay
Brand elements — Logo visibility, brand color consistency

Plus cross-competitor visual comparison identifying visual whitespace opportunities.

What this replaces

Before	After
Browsing Facebook Ad Library manually	Automated scraping and downloading via Glued MCP
Screenshots in a Google Drive folder	Structured image downloads with metadata
"Their ads look nice" level analysis	Per-image visual breakdown across 5 dimensions
Gut-feel creative briefs	Data-driven briefs targeting specific competitive gaps
Generic hook brainstorming	15 typed hooks per competitor with strategic rationale
No competitive dashboard	Interactive Chart.js dashboard with 6 panels

Who this is for

Creative strategists who want visual competitive intelligence, not just copy analysis
Performance marketers building creative briefs grounded in competitive data
Agency teams running competitive audits with deliverable-ready output
Brand managers tracking competitor visual identity and messaging evolution
Founders entering a market who need to understand the creative landscape fast

Requirements

Requirement	Details
Glued MCP	Access to `scrape_meta_ads_library` tool
Claude Code	Runtime environment
Node.js	For dashboard generation script
curl	For downloading ad images (standard on macOS/Linux)

No API keys or environment variables required.

Skill structure

competitor-research/

├── SKILL.md # Skill definition with 4-phase workflow ├── references/ │ └── subagents.md # Full prompts for Recon, Analyst, and Strategist agents ├── scripts/ │ └── generate-dashboard.mjs # Zero-dependency HTML dashboard generator └── README.md # This file

How it works under the hood

Agent	Model	Job
Recon	Haiku	Scrape Meta Ads Library, filter to IMAGE, download creatives
Analyst	Sonnet	Quantitative + visual creative analysis using Claude vision
Strategist	Sonnet	Creative briefs, hook generation, variation matrix
Dashboard	Node.js script	Interactive Chart.js HTML (no LLM cost)

Key technical details:

Recon filters to display_format: "IMAGE" — drops VIDEO and CAROUSEL entirely
Analyst uses Claude's native Read tool on image files for zero-cost visual analysis
Dashboard script has zero npm dependencies — uses Chart.js via CDN
Supports both page_ids and search_keywords modes for scraping
Brand voice file auto-detected from common paths for Strategist output

For marketers who believe competitive intelligence should include what ads look like, not just what they say.