Maintain automations

1. Find and read: search(workspace_id, query), list_workflows, list_checklists. get_workflow(workflow_id) shows steps, connections, execution_order and unreachable_step_ids; read it before changing anything.
2. Trigger and details: update_workflow(workflow_id, name?, description?, trigger_config?). For a schedule send { cron, timezone } with the timezone from get_me.
3. Steps:
   • update_workflow_step(step_id, config) replaces the config in full: send every key.
   • add_workflow_step(...) appends and connects; remove_workflow_step(step_id) repoints its inputs at its outputs.
   • reorder_workflow_steps(workflow_id, step_ids) sets one linear order. Pass every id (one left out is disconnected); it flattens branches.
4. Branches: connect_workflow_steps(workflow_id, source_step_id, target_step_id, branch). On a condition step branch is 'true' or 'false'; otherwise null. A step appended after a condition lands on true; false is unconnected by default, so a failed condition ends the run unless you connect it. source_step_id: null is the trigger; target_step_id: null disconnects.
5. Checklists: get_checklist(checklist_id), then update_checklist(checklist_id, checks=[…]) sending each existing check with its id. A kept id keeps run history; a dropped check is deleted with its results. Also adjustable: instructions, url, actor_id, pass_threshold, run_mode.
6. Verify: get_workflow again (all steps in execution_order, none unreachable), then run_workflow or run_checklist once.
7. Pause or retire: prefer toggle_workflow to pause. delete_workflow and delete_checklist discard all run history, so use them only when the user asks.

Example, notify only on failure: add_workflow_step(workflow_id, step_type="condition", config={ field: "previous_step.failed", operator: "gt", value: 0 }), then add_workflow_step(workflow_id, step_type="notification", config={ channel: "email", recipients: { type: "all_members" }, template: "failed" }).