Back to Portfolio
April 20266 min read

I Built a Chrome Extension That Lets AI See Your UI — Here Is Why It Matters

SS

Shrikant Satpute

Software Engineer

Every frontend developer knows the pain: you spot a UI issue, switch to your AI assistant, and spend five minutes describing which button, which padding, which hover state needs fixing. The AI guesses wrong, you correct it, and three rounds later you have wasted more time than just fixing it yourself. I built Shadowmark to eliminate that entire loop — a Chrome extension that gives AI agents visual context about your UI.

The Problem: Lost in Translation

When working with AI coding assistants on frontend tasks, the biggest bottleneck is not the AI's coding ability — it is context transfer. Consider this typical exchange: "The submit button on the third card in the dashboard grid has wrong padding and the hover color is off." The AI has to guess which component file, which CSS class, which element you mean. It often gets it wrong. You spend more time describing the problem than fixing it. I realized the missing piece was visual context — the AI needed to see what I was seeing.

How Shadowmark Works

Shadowmark is a free Chrome extension with a dead-simple four-step workflow: 1. Point — Hover over any element on any website. Shadowmark highlights it with a visual overlay. 2. Click — Select the element. Shadowmark instantly captures its CSS selector, XPath, computed styles, ARIA attributes, and framework-specific data (React props/state, Vue setup data, Angular component info). 3. Describe — Add a note explaining what you want changed: "increase padding to 16px" or "change hover background to brand blue." 4. Ship — Export everything as structured Markdown or JSON, paste it into Claude Code, and watch the AI make the exact change you described. The entire toolbar lives inside a Shadow DOM, so it never interferes with the host page styling.

Framework-Aware Intelligence

What makes Shadowmark powerful for enterprise projects is its framework awareness: • React: Extracts component names, props, and state from the React DevTools fiber tree • Angular: Identifies component classes and template bindings • Vue: Reads setup data and reactive properties • Plain HTML/WordPress: Falls back to standard DOM inspection This means when you export an annotation, the AI does not just get a CSS selector — it gets the actual component name and prop values. Claude can jump directly to the right file and make surgical changes.

Real-World Impact

Since building Shadowmark, my frontend debugging workflow has transformed: • UI bug reports that took 5-10 minutes to describe now take 30 seconds • AI makes the correct fix on the first attempt 80%+ of the time • Multi-element selections let me batch 5-6 UI tweaks into a single AI prompt • The Markdown export format is optimized for AI parsing — no wasted tokens Shadowmark is free, open source, requires no signup, and has zero tracking. It works on any website — even sites you do not own — making it useful for competitive analysis and design inspiration too.

Shadowmark Workflow

Point & Hover
Click & Capture
Describe Change
Export to AI
Back to Portfolio

Written by Shrikant Satpute