Problem\n\nWhen processing very small audio files (e.g., 22KB), the whisper transcription attempt fails and wastes resources. The files are often silent, corrupted, or too short to contain meaningful audio.\n\nExample: A 22KB OGG file produced empty transcription output despite the skill attempting to process it.\n\n## Proposed Solution\n\nAdd a pre-processing check to automatically skip files that are:\n- Below a size threshold (e.g., < 50KB)\n- Below a duration threshold (if duration can be quickly determined)\n- Or have other indicators of being non-viable for transcription\n\n## Expected Behavior\n\nInstead of attempting transcription and failing, the skill should:\n1. Check file size/duration before starting Whisper\n2. If below threshold, return early with a clear message like "Audio file too small/short to transcribe"\n3. Skip to the next task without wasting resources\n\n## Why This Matters\n\n- Saves API/compute resources\n- Provides clearer feedback to users\n- Avoids confusing empty output files (.txt, .srt, .vtt all created as empty files)\n\n---\n\nThis issue was raised by Martin via OpenClaw.