trapstreet/trapstreet-tasks

Open evaluation tasks for AI agents, skills and tools — inputs, expected outputs and a judge, so any solution can be measured and ranked on trapstreet.run