arihantprasad07/guni ? reverse-engineered prompt
Reverse engineered prompt
Build me a Python security tool for agentic browsers that checks a web page before an AI agent acts on it.
I want it to scan raw HTML and catch things like prompt injection, phishing forms, clickjacking, hidden overlays, redirect tricks, malicious scripts, and goal hijacking, then return a simple decision like allow, confirm, or block with a risk score and evidence. Please include a small Python SDK, a FastAPI service with a scan endpoint that can be protected by an API key, and a basic dashboard so I can see results and audit what happened. It should also work with browser automation flows like Playwright and be easy to call from other agent tools.
If it helps, add an optional LLM based reasoning mode for deeper checks, but keep the default fast and reliable. Make it runnable locally with clear setup instructions, tests, and a clean demo page. If you need to look up current docs online, go ahead.
Are you gonna build this?
make sure you review the code using coderabbit