What's this about?
WeClawArena is an auditable sandbox and benchmark for networks of personal agents — where each user has an AI acting on their behalf and those agents must collaborate with, and stay secure against, each other.
What exactly is a "human-centered agent network"?
Picture everyone having a personal AI that books, negotiates, and coordinates on their behalf. Those agents then have to talk to *other people's* agents. That's an agent network — and it creates new security problems: one user's agent might try to manipulate or extract data from another's. WeClawArena is a controlled arena to study and audit exactly those interactions.
Why it matters
It's directly relevant to today's trace-stealing story: as agents act for us and talk to each other, the security surface explodes, and we need testbeds to measure it before deploying at scale.