[Dummy] Tool response type export for codegen

Response for find_agent tool.

Response when no agents found.

Response for get_details action.

Response for validate action.

Response for run/schedule actions.

Response when login is needed.

Response for errors.

Response when run_agent receives unknown input fields.

Response for agent_output tool.

Response for add_understanding tool.

Response for previewing a generated agent before saving.

Response when an agent is saved to the library.

Response when the LLM needs more information from the user.

Response when the goal needs refinement with a suggested alternative.

Response for find_block tool.

Response for block details (first run_block attempt).

Response for run_block tool.

Response for search_docs tool.

Response for get_doc_page tool.

Response when MCP tools are discovered from a server (agent-internal).

Response after executing an MCP tool.

Response for decompose_goal tool — shows the plan to the user.

Response containing a list of schedules.

Response confirming a schedule was deleted.

Response listing the user's presets.

Response confirming a preset was updated.

Response confirming a preset was deleted.

Response when a memory is stored.

Response when memories are searched.

Response with candidate memories to forget.

Response after deleting specific memory edges.

Ack returned by TodoWrite.

The tool is effectively stateless — the authoritative task list lives in the assistant's latest tool-call arguments, which are replayed from the transcript on each turn. The tool output only needs to confirm that the update was accepted so the model can proceed.