TrackMyWork

Inspiration - The modern web is a minefield. Every creator, researcher, and casual user is constantly tracked by hidden scripts, analytics beacons, and manipulative consent flows.

What inspired us was the realisation that most people don’t even know when their data is being quietly exploited. We wanted to give users back visibility and control, without them needing to read hundreds of lines of privacy policy text.

What it does - TrackMyWork is a Chrome Extension that scans any webpage in real time and detects signs of data exploitation.

It combines Chrome’s built-in AI APIs (Summarizer + Prompt) with a heuristic keyword scanner.

How we built it - We used Manifest V3 to define the extension.

• Content scripts extract page text and split it into sentences.
  
• Background service worker orchestrates AI calls via Chrome’s window.ai (Prompt + Summarizer APIs).
  
• Popup UI built in HTML/CSS/JS renders the risk meter, evidence list, and advice.
  
• If AI is unavailable, we fall back to a heuristic keyword model (e.g. track, cookie, analytics, fingerprint).
  
• Export is handled through chrome.downloads.download() producing JSON reports like:
  

{
  "risk_level": "High",
  "evidence": ["cookie tracking", "analytics beacon"],
  "confidence": 0.6,
  "advice": "This site may track you. Consider blocking trackers."
}

## Challenges we ran into - AI integration hurdles: Chrome’s new `window.ai` APIs are experimental, and injecting them safely into pages without breaking manifest v3 rules was a major challenge.  
- Fallback design: We had to ensure TrackMyWork never returned an empty verdict. Crafting a reliable keyword-based heuristic fallback required testing across very different websites.  
- UX trust gap: At first, a single word verdict like “Low” felt fake. We needed a risk meter, evidence list, and advice box so users could *see* the analysis behind the answer.  
- Theme visibility: Our dark mode initially hid the risk meter text — fixing it required dynamic theming.  

## Accomplishments that we're proud of - Built a fully working Chrome Extension in v6 that scans any website for exploitative data practices.  
- Integrated AI-powered summarisation and prompt analysis directly inside the browser with no backend.  
- Designed a risk meter UI with dark/light themes, evidence highlights, and exportable reports.  
- Proved real functionality with contrasting outputs: one “Low risk” report and one “High risk” report, showing it adapts dynamically.  
- Created a solution that is lightweight, privacy-first, and network-resilient.  

## What we learned - How to leverage Gemini Nano + Chrome AI APIs for client-side intelligence.  
- The art of writing strict JSON-output prompts for structured, reliable results.  
- The value of fallback resilience: users must get an answer even if AI is offline.  
- That visible evidence and advice builds trust in AI systems — transparency matters more than just verdicts.  
- How design polish (risk meters, dark mode, smooth export flows) turns a raw tool into something judges see as real.  

## What's next for TrackMyWork - Mobile extension: scan data misuse in in-app browsers on Android and iOS. 

Built With

• css3
• html5
• javascript
• local
• manifest
• prompt
• scripting
• storage
• summarizer