ViewDAO
📉 OpenClaw has 135,000 GitHub stars and a critical problem: 41.7% of its agent skills contain serious security vulnerabilities.
Researchers just found 26 LLM routers secretly intercepting agent commands. One drained $500K.
The race to ship agentic AI has outrun the ability to secure it. But we may have the solution to solve this soon 🔥
Researchers have discovered that 41.7% of the agent skills of OpenClaw (135,000 GitHub stars) have serious security vulnerabilities, with 26 LLM routers secretly intercepting agent commands, one of which resulted in a loss of $500,000. The AI agent race has surpassed security capabilities, but a solution may be available soon.