Safety benchmark for personal AI agents under realistic prompt injection. 120 test cases across 5 professional domains, 3 injection vectors, and 5 harmful action types. (arXiv 2026)
Overview
ClawSafety (Benchmark) is a safety & security AI agent developed by weibowen555 with 20 GitHub stars, written in Python, currently Archived.
Safety benchmark for personal AI agents under realistic prompt injection. 120 test cases across 5 professional domains, 3 injection vectors, and 5 harmful action types. (arXiv 2026)
Retired after prolonged inactivity · 20 stars at retirement. It returns to the directory automatically if its repository becomes active again.
GitHub metrics refresh daily — last fetched 20d ago.