A minimal, scenario-driven privacy assistant built with Claude 3.5 prompts. DataGuardian helps users assess digital risks, understand privacy threats, and take safer actions — all through mapped, modular AI coaching.
Try the app here:
https://partyrock.aws/u/prithika/1IfgddUWL/DataGuardian%253A-AI-Powered-Privacy-Coach
- Claude 3.5 Sonnet (
bedrock-anthropic.claude-3-5-sonnet-v2-0) - Zero temperature for consistent, mapped output
- Modular widget layout with prompt-driven logic
| Widget Title | Functionality |
|---|---|
| Static Text | Intro message: “Worried about your digital footprint?” |
| Your Privacy Scenario | Text input for describing a concern or situation |
| Risk Detector | Categorizes risk (e.g. Phishing, Data Leak, Over-Sharing) + short reason |
| Ask DataGuardian | Friendly, actionable advice with checklist-style guidance |
| Privacy Risk Badge | Returns a visual badge (Low 🟢, Medium 🟡, High 🔴) based on risk level |
The full widget configuration is stored in layout.json.
Includes:
- Widget coordinates and dimensions
- Prompt logic for each module
- Claude model parameters
See /screenshots for UI previews:
interface-1.png: Full layout viewinterface-2.png: Module breakdown
- Privacy education
- Scenario-based coaching
- Cybersecurity awareness
- Portfolio demonstration of AI prompt engineering
This project follows a mapped, modular design:
- Input → Risk Categorization → Advice → Badge
- Each module runs independently and returns deterministic output
- Ideal for showcasing prompt clarity and AI coaching logic