Visual Research

Deep brand intelligence, automated.
Built by Curious Endeavor.

 
--- Analyze any brand's visual identity. A Claude Code skill by [Curious Endeavor](https://curiousendeavor.com). --- A five-phase pipeline that discovers agency work, captures social presence, extracts images from APIs, runs vision analysis, and packages everything into a branded 13-section HTML report. Built for brand strategists, creative directors, and anyone preparing competitive intelligence on a brand's visual identity. ## Examples Reports generated by the skill for real brands:

Coca-Cola — hero section with TL;DR executive summary

Spotify — positioning section with Wrapped 2025 imagery

Cash App — digital experience section with homepage screenshot

Wise — brand evolution timeline and brand system imagery

Venmo — social and campaigns section with 'What's Your Venmo?' campaign

## What it does - **Discover** — Find agency relationships, campaign case studies, and press coverage via web search - **Capture** — Screenshot websites, social grids, and app stores with Playwright - **Extract** — Pull actual images from Instagram API, YouTube, Prismic CMS, and agency portfolios - **Analyze** — Run vision analysis on captured imagery (Gemini or Claude) to extract colors and composition - **Package** — Generate a structured markdown document and a branded HTML report covering 13 sections of competitive analysis ## Quick start ```bash # 1. Clone git clone https://github.com/curious-endeavor/visual-research.git # 2. Copy to your Claude Code skills directory cp -r visual-research ~/.claude/skills/visual-research # 3. Open Claude Code and say: # "research Cash App" # "visual research on Stripe" # "analyze Nike's brand identity" ``` ## Report sections The HTML report includes a hero with TL;DR executive summary, then 12 sections: ``` §01 Quick Facts §07 Assessment §02 Positioning §08 Strategic Verdict §03 Brand Identity §09 Brand Personality §04 Brand Evolution §10 Social Strategy §05 Brand in Practice §11 Campaigns §06 Digital Experience §12 vs [Client] ``` Plus a full-width "Signal" dark break between §03 and §04 highlighting the single most important strategic insight. ## Pipeline ``` DISCOVER ─── find agencies, campaigns, press │ CAPTURE ──── screenshot websites, social grids, app stores │ EXTRACT ──── pull images from Instagram API, YouTube, agency sites │ ANALYZE ──── vision analysis → colors, type, composition │ PACKAGE ──── markdown doc + branded HTML report ``` ## Output The skill produces two files: 1. **`[brand]-visual-research.md`** — Structured markdown with all 13 sections, suitable for feeding into other tools 2. **`[brand]-visual-research.html`** — Self-contained branded HTML report that opens in your browser The HTML report follows the [Curious Endeavor styleguide](https://www.curiousendeavor.com/styleguide.html) — Larken headlines, Inter body text, JetBrains Mono for technical labels, #CC0000 red accent used sparingly. ## Customization The report template uses CSS variables. Change the branding by editing the `:root` block in `templates/report.html`: ```css :root { --red: #cc0000; /* Accent color */ --black: #1a1a1a; --grey: #666; --light: #999; --border: #eee; --bg: #fff; --max-w: 1100px; /* Content max width */ --font-display: 'larken', 'DM Serif Display', Georgia, serif; --font-body: 'Inter', -apple-system, sans-serif; --font-mono: 'JetBrains Mono', monospace; } ``` Larken loads from Adobe Fonts (Typekit). If you don't have a Typekit subscription, DM Serif Display (Google Fonts) kicks in automatically as the fallback. The CE logo is base64-embedded in the template. Replace the `` src in the `.ce-bar` to use your own logo. ## Trigger phrases Say any of these to invoke the skill: - "research [brand]" - "visual research on [brand]" - "brand visual audit for [brand]" - "analyze [brand]'s brand identity" - "competitor visual analysis of [brand]" ## Prerequisites **Required:** `curl` (pre-installed on macOS/Linux) **Recommended:** | Tool | What it unlocks | Install | |------|----------------|---------| | `GEMINI_API_KEY` | Vision analysis of captured images (Phase 4) | [Google AI Studio](https://aistudio.google.com/) | | Playwright | Full-page screenshots (Phase 2) | `npx playwright install chromium` | | yt-dlp | YouTube/TikTok thumbnails (Phase 3) | `pip3 install yt-dlp` | The skill works without any of these — it falls back to web search and Claude's built-in vision. Each tool adds richer output. ## File structure ``` visual-research/ SKILL.md # The skill (Claude reads this) templates/ report.html # HTML report template (full-width, pill nav) assets/ ce-logo-red.png # CE logo source (embedded as base64 in template) references/ brand-audit-framework.md # 13-section research template + HTML mapping source-discovery-guide.md # Where to find brand assets image-extraction-techniques.md # Instagram API, Prismic, yt-dlp recipes ce-styleguide.md # Curious Endeavor design rules ``` ## About [Curious Endeavor](https://curiousendeavor.com) is a brand agency that uses agentic branding flows. This is our first open-source skill — built from the same pipeline we use for competitive research on real client work. ## License [MIT](LICENSE)