ai-agents-multi-action-agent-r-multi-action-agents-raise-a-valueerror-when-any-to-6b3160d3

THEN Remove or modify the `validate_return_direct_tool` validator in `AgentExecutor` to allow `return_direct=True` in multi-action site_1. When a tool with `return_direct=True` finishes, force the agent to return its output directly and stop further tool calls.

In langchain/agents/agent.py, remove the validation block that raises `ValueError` when a tool in a multi-action agent has `return_direct=True`. Then update the agent execution logic to check `tool.return_direct` after a tool call and, if true, break the loop and return the result.