WeClawArena: An Auditable Sandbox and Benchmark for Cross-User Agents Collaboration and Security in Human-Centered Agent Networks

2026-08-04

AI Summary

This work addresses the limitations of existing agent benchmarks, which lack support for verifiable cross-user collaboration and security auditing, and fail to model multi-agent interactions and malicious behavior propagation in realistic digital workspaces. To bridge this gap, we present the first end-to-end auditable sandbox platform centered on human-agent networks. By integrating sandbox isolation, fine-grained logging, and multi-role workspace modeling, our platform enables users to deploy their own agents in private environments while fully tracing communications, tool invocations, and resource operations. The framework incorporates benign controls and four attack vectors, automatically generating a comprehensive benchmark spanning six domains, 124 core tasks, and 620 scenario variants. This benchmark simultaneously evaluates task utility and attack success rates, effectively diagnosing issues such as collaboration failures, privacy leaks, and privilege escalations.

Abstract

Recent advances in persistent personal-agent frameworks are making human-centered agent networks realistic deployment targets: each user can be served by an AI agent that acts on the user's behalf, maintains state, and communicates with other agents through social and task relations. In these networks, everyday tool use becomes multi-party owned-agent collaboration over personal workspaces, where files, records, tools, and policies are not directly visible across owners. Existing agent benchmarks study tool use and collaboration, but they do not provide an end-to-end sandbox for verifiable cross-user agent collaboration with realistic user digital workspaces or test how harmful actions can travel through the human-centered agent network. We introduce WeClawArena, an auditable benchmark and runtime sandbox for multi-party owned-agent collaboration over personal workspaces. WeClawArena targets collaborative tool-use tasks in which personal workspaces serve as both operational tools and personal constraints. The benchmark contains 124 base tasks across six cross-user task domains and expands them into 620 scenario variants, with one benign control and four attack-vector variants per base task. The sandbox records peer messages, tool calls, resource operations, governed decisions, and final workspace states. WeClawArena reports utility and attack success rate separately and audits attack success from bounded runtime evidence, supporting diagnosis of task breakdown, privacy leakage, poisoned evidence, and invalid authority paths.

Problem

• cross-user agent collaboration
• human-centered agent networks
• auditable sandbox
• security
• personal workspaces

Innovation

• auditable sandbox
• cross-user agent collaboration
• human-centered agent networks
• attack vector benchmarking
• personal workspaces