Feature request: Support trusted, non-LLM-controlled metadata inputs for Agent tools

Is your feature request related to a problem? Please describe. In the new n8n Agents / Agent Builder experience, tool inputs appear to be driven by the agent's generated arguments. There does not appear to be a native way to provide separate trusted metadata values to a tool that are resolved by the workflow/runtime and cannot be generated or altered by the LLM. This is important for workflows where a tool needs application context such as customer_id, tenant_id, session_id, authorization scope, or values produced by an earlier workflow/tool. These values should not need to be exposed as agent-selectable parameters. Describe the solution you'd like Add a separate context/metadata input mechanism for Agent tools, distinct from the tool-call argument schema. For example: • Agent-controlled arguments: order_status, selected by the model based on the user's request. • Trusted runtime context: customer_id, tenant_id, and session_id, supplied by the parent wo…