Skip to main content
Gate dangerous tool executions with ToolContext.request_confirmation(): 
from langchain_adk.tools import ToolContext

async def confirm_dangerous(ctx, tool_name: str, tool_args: dict) -> None:
    if tool_name in ("delete_file", "drop_table"):
        tool_ctx = ToolContext(ctx)
        tool_ctx.request_confirmation()
When request_confirmation() is called from a before_tool_callback, the tool execution is paused and a confirmation request event is yielded. The caller must approve or reject before execution continues. 
agent = LlmAgent(
    name="SafeAgent",
    llm=llm,
    tools=[delete_file, drop_table, search],
    before_tool_callback=confirm_dangerous,
)