Null task_input_files for 34 tasks with empty input snapshots (inputs are world-level)

#15

Summary

Sets task_input_files to null for the 34 tasks whose referenced input snapshot is empty (0 files), making the field accurate and the dataset self-consistent.

Background

175 tasks had a non-empty task_input_files (a snap_* id), but only 141 have a corresponding task_files/<task_id>/ directory. For the other 34, an input snapshot was attached during authoring but never populated — these tasks operate on world-level files already shipped in world_files_zipped/<world_id>.zip (verified, e.g. World226_TA_01 → PLTF_LBO_v06.xlsx; World227_SK_Task06 → MFC Model_v4.xlsx; World246 → KVUE/DCF models). An empty input snapshot is functionally identical to null.

Change

  • task_input_files: "snap_…"null for the 34 affected tasks (diff is exactly 34 lines).
  • Result: every non-empty task_input_files (now 141) has a matching task_files/<task_id>/ directory — no dangling references. No files are removed; these tasks remain solvable from the world filesystem.

Not changed

The 141 populated input sets, gold files, rubrics, and all other fields are untouched.

jwstanly-mercor changed pull request status to merged

Sign up or log in to comment