$ python3 -I tools/experiments/characterbench_prompt_check.py --compare <bundle>/results/prompt_hashes_authors.json <bundle>/results/prompt_hashes_platform.json
# started 2026-10-27T14:03:10+00:00  ended 2026-10-27T14:03:10+00:00  exit 0

samples: 6494 / 6494; only in first: []; only in second: []
role-play prompt mismatches: 0 []
judge prompt mismatches: 0 []