[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.

Response for the two-step Soul edit tools.

Carries the diff so the model must surface exactly what changed. applied is False for the update_expert_soul preview (nothing written yet; the one-time confirmation_id references the stored proposal) and True once confirm_expert_soul_update saves the edit.