Search for a command to run...
LongTraceRL: Learning Long-Context Reasoning from Search Agent Trajectories with Rubric Rewards