• 

  App Interface

• 

  SPAM Mail

• 

  Pasted Spam Mail

• 

  Threat Assessment Report (1/2)

• 

  Threat Assessment Report (2/2)

Scam Detective ∙ AI Safety Companion

What problem your app solves and why we built it

Digital scams, phishing campaigns, and sophisticated social engineering tactics are skyrocketing. While traditional spam filters catch generic threats, they fail against hyper-targeted, text-based manipulation - like fake LinkedIn job offers, urgent SMS warnings, or spoofed invoice requests. These scams succeed not through malware, but through psychological manipulation.

We built Scam Detective to serve as an instant AI Safety Companion. It gives everyday users a secure sandbox to drop any suspicious message into and receive an immediate, visual breakdown of the hidden threats. Instead of a simple "safe/unsafe" verdict, it strips away the attacker's leverage by exposing their psychological triggers, saving users from devastating financial and data losses.

How we structured conversations with MeDo to build our project

To build this application entirely during the hackathon window, we leveraged MeDo’s Deep Build feature with a highly structured architectural blueprint:

1. Schema and Core State: We initiated the project by prompting MeDo to establish a robust backend relational database (scans and remediations tables) to safely log submissions and threat categories.
2. Component Isolation: Once the backend foundation was set, we used iterative, multi-turn logic to construct the reactive UI, ensuring seamless state transitions between the input sandbox and the final dynamic threat report.
3. Refining the Analysis Engine: We instructed MeDo to handle raw text payloads through an advanced extraction engine, translating unstructured text blocks into highly structured JSON data fields that feed our frontend visual meters.

The most impressive feature MeDo helped us create

The absolute best part generated by MeDo is the "Deconstructed Payload" Panel on the main report dashboard.

Instead of throwing a flat block of text or a boring summary back to the user, MeDo perfectly built an interactive, highlighted text block. It intelligently parses the pasted scam message, isolates specific substrings, and applies distinct visual highlight tints to the text in real-time. For example, if a message uses fake urgency like "Your account will be suspended in 2 hours," MeDo highlights it in high-contrast red. When a user hovers over the highlighted text, an inline tooltip smoothly reveals the exact psychological lever being pulled by the attacker (e.g., Fear/Urgency Tactic).

The level of UI polish, coupled with complex token-matching backend logic generated on the first try, completely blew us away and makes the application look like an enterprise-ready security asset.

Built With

• medo
• supabase