I just saw that Yu Xian from SlowMist shared his concerns about OpenClaw on X. Basically, he says that although both tools have security mechanisms, there is a real problem with OpenClaw: the excessive freedom it offers makes it difficult to maintain proper stability control.

What’s interesting is that he identifies a clear contradiction: OpenClaw aims to be more open and granular in permissions, but that same openness makes it less controllable. He mentions that in real production environments, users constantly face this tension between openness and security. With Claude Code, the situation is different; it builds more trust precisely because it prioritizes control.

Although OpenClaw also has sandboxing, Yu points out that certain versions are simply not sufficiently committed to stability. An interesting perspective considering that security should be fundamental in any development tool. Do you already use any of these?