[jmjava]

Summary

• Post-tts lint used to pass a Whisper transcript that is empty or whitespace, so audio with no spoken words cleared the artifact scan.
• An empty transcript is now a lint failure, matching the pre-tts rule that empty narration is not spoken text. The proving tests fail if that check is removed.

Test plan

• tests/test_narration_lint.py with PYTHONPATH pointed at the worktree
• CI lint, unit, benchmark, validate, mutation, and GitGuardian

Made with Cursor