uefn-create-use-remove-one-script

Rule

Create + use + read back + cleanup in ONE ProgrammaticToolset/execute_tool_script, because:

1. the guard allows writes only to objects created by the same process;
2. errors from internal tools cannot be caught in the script and there is no rollback: a script that aborts leaves what it created, so the CALLER cleans up after a failed script (AssetTools/delete, SceneTools/remove_from_scene);
3. exists right after delete may return true (materials trap 12): verify with find_assets o exists in .Name form. Re-read every write: add_rows, set_rows, add_tag return null, not true.

Verified recipes

• execute_tool_script#create_widget_chain (CreateWidgetBlueprint, AddWidget, ReplaceWidgetWithTemplate, set_properties, get_properties; cleanup AssetTools/delete of the folder {lab_mount}/...).
• execute_tool_script#batch_reads (multiple reads in one call).
• execute_tool_script#verse_edit_build.
• Script shape: def ex(n,a): return execute_tool(n, json.dumps(a))["returnValue"] e def run(): that returns a dict.

Paths

Always GUID: {lab_mount} = /<guid>/<project>. values of set_properties is a JSON STRING with camelCase names; delete takes the path without .Name.

Don't

Leaving an actor in the level (a leftover T008_Cube cost a whole task); placing actors that simulate physics; run() returning a non-dict; bypassing a guard refusal; improvising if it fails (exact error, caller cleanup, stop).

To try (not verified)

• recipe/apply_material_instance_to_actor (ONE script): unverified.
• collision_set_and_revert: unverified (reset_properties goes to the class default, not the original: explicitly restore the original values).
• recipe/animate_owner_actor_with_script_and_restore: unverified.