SensHawk

The document is pasted into ChatGPT and then the extension scans the document for sensitive information before being sent.
A screenshot in the ChatGPT homepage showing the extension running in the top right and the fixed prompt after removing sensitive info.

Inspiration

After brainstorming several ideas that would help us get tasks finished from as many sponsors as possible, we decided on developing a simple, lightweight, and open source document scanner.

What it does

This privacy scanner intercepts any prompts before it is sent into the AI, and if sensitive information is detected once it is sent to the back-end, it'll stop the prompt and notify you about the sensitive information being leaked, as well as offer you a sanitized version you can optionally sent instead.

How we built it

The project was split 3 ways, 2 beginner developers and one experienced developer. One worked on the UI using HTML and CSS, One worked on the Backend using Python, Flask, Jsonify, and Request, and One worked on the Frontend using JS, Css, Json, Python, and HTML.

Challenges we ran into

One beginner developer had a bit of trouble working with new tools such as Regex, Jsonify, and Flask. As well as using Git, meanwhile for the front-end more experienced developer, it was still quite difficult dealing with advanced DOM manipulation (and the shadow DOM) to be able to intercept the AI prompt.

Accomplishments that we're proud of

We were able to get the project up in runnning in about a day's time. As well as manage our schedule to centralize the project.

What we learned

We learned how to use new tools and libraries to get exactly what we wanted and not only simplify work but also optimize it.

What's next for SensHawk

We plan to continue support for SensHawk to continue adding features

Built With

css
flask
html
javascript
json
jsonify
pycharm
python
regex
request
vscode

Submitted to

ShellHacks 2026

Created by

I designed the extension’s user experience in Figma, including the personal information warning and the suggested redacted prompt. I also tested the Chrome extension locally, confirmed that it detected a test email address, and checked that the sanitized prompt could be sent.

Sofia Ouanter
I worked on the front-end of the extension, mostly focusing on the Javascript for intercepting the prompt. I also helped with teaching a lot of the basic technologies used during these events, like git, among taking other roles.

Christopher Abreu
I worked on the bacjend of the projecett using Pythoh, Flask, Jsonify, and Request. I am a beginner so getting used to these libraries and the tools provided for the first time was a bit rough but after reading the docuentation and viewing examples of each library being used helped me learn how to use each tool. In the end i was able to finish my side of the project and get the backend up and running.

Luis Cruz