A crypto streamer lost $100K because his seed phrase was visible on a live stream for a few seconds. I watched the clip and thought — why doesn't a tool exist that catches this automatically?
So I built one with Claude Code.
ScreenCloak watches your screen using real-time OCR while you stream. When it detects sensitive data, it auto-switches OBS to a safe scene before viewers see it.
What it catches:
BIP-39 seed phrases (2048-word list matching)
Credit card numbers (Luhn validation)
Crypto addresses (BTC, ETH, SOL)
API keys (AWS, GitHub, Stripe, OpenAI + 10 more)
Personal strings you define
I'm not an engineer — I'm a product person learning to build. Claude Code wrote the code, but the problem discovery, product decisions, and testing were all me. And honestly, the testing was the hardest part. Turns out common English sentences can trigger seed phrase detection, and random numbers in URLs look like credit cards to a regex.
It's free, open source, and I'm looking for feedback.
🎥 Demo https://youtu.be/IL_hvf-CLek
💻 GitHub https://github.com/tomershahar/ScreenCloak
🔗 Site https://screen-cloak.base44.app/
💬 Discord https://discord.gg/e46csQsuRc
Tags: #ai #claude #opensource #buildinpublic
Top comments (0)