Inspiration
Regular AdBlock's use blocklists, lists of known URL's, server addresses, and/or file hashes, to block advertisements and malware. These blocklists can miss harmful advertisements when blocking, we were sick of misses so we build Apollo AdSheild.
What it does
Apollo AdSheild is a Chrome extension that finds ads on a web page and decides whether each one is safe. Blocklists only stop ads that someone has already reported. Apollo AdSheild uses an AI model to judge ads that aren't on any list, such as fake "Your Chrome is out of date" warnings and deceptive sponsored cards.
How we built it
We used Opencode (models big pickle and Claude haiku 5.5) to do the actual programming, but our team came up with the project structure and logic behind it.
Extension: Chrome Manifest V3, plain JavaScript with no build step. A content script scans every frame, hides candidates, and applies verdicts. A background service worker records redirect chains and talks to the classifier. The popup renders the report.
Classifier: a small Node.js server with no dependencies. It sends candidates to any OpenAI-compatible model in JSON mode, validates the response against a fixed schema, and falls back to local rules if the model fails. It can also check URLs against Google Safe Browsing.
Verdicts: a threat taxonomy (malicious, deceptive, phishing, unwanted redirect, tracker, benign ad, legitimate) with severity and an action (block, warn, or allow). Each verdict includes human-readable indicators.
Demo: a fake news site with a malicious fake-update ad, a sponsored card, and a privacy notice that must not be hidden, plus a harmless phishing mock page.
Challenges we ran into
Surprisingly, the development process was rather smooth! we ran into a total of 0 bugs and 0 challenges with development! our planning stage lasted about an hour, and after we fed our plans for the project to Opencode, the full extension was completed in less than 30 minutes.
Accomplishments that we're proud of
Our team fully finished our project in less than 2 hours. We worked efficiently with each other and our tools to quickly create something helpful.
What we learned
We learned a lot about the advancement of ai on this project, none of us had really used ai up to this point. We learned how powerful these tools will be for our future careers and how important it is that we master them.
What's next for Apollo AdSheild
Apollo AdSheild is a fully finished project. While it was fun to develop at plutohacks 2026, no more work will be done on Apollo AdSheild after plutohacks 2026. We believe Apollo AdSheild is perfect just the way it is!
Built With
- css
- google-safe-browsing-lookup-api
- html
- javascript
- manifest-v3
- node.js
- opencode
- powershell
- python
- visual-studio
- winget