agentbench

// Benchmark your OpenClaw agent across 40 real-world tasks. Tests file creation, research, data analysis, multi-step workflows, memory, error handling, and tool efficiency. Not a coding benchmark — measures your agent setup and config.