OpenClaw Security Testing: 80% hijacking success on a hardened AI agent

(earlycore.dev) by paldeac2 | Feb 6, 2026 | 0 comments on HN