| description | Generate a cut-for-time report from prepared transcript artifacts |
|---|
Generate a cut-for-time report for: $@
Workflow:
- Identify the target episode slug and the matching transcript directory or prepared transcript artifacts.
- If
dist/analysis/<slug>/transcript_chunks.mdordist/analysis/<slug>/transcript_index.jsondo not exist yet, run:uv run python scripts/prepare_transcript_analysis.py <transcript-dir> --output-dir dist/analysis/<slug> --slug <slug> --plain-output-names - Read the prepared transcript artifacts. For long transcripts, use the chunk index first and then inspect the weakest or most repetitive chunk sections.
- Use pi-driven judgment to identify sections that can be trimmed or cut.
- Write the final report to
dist/analysis/<slug>/cut_report.md. - Reply with the output path and a short summary of the biggest cut opportunities.
Output expectations:
- include exact time ranges for each suggested cut or trim
- explain why each section is weak, repetitive, speculative, low-energy, or too niche
- estimate likely time savings when possible
- distinguish between hard cuts and lighter trims when useful
- mention especially strong sections that should definitely stay if that helps the edit
Quality bar:
- optimize for a tighter, stronger episode rather than maximizing total cuts
- avoid cutting sections only because they are technical if they still have clear payoff
- prefer cuts that reduce repetition, dead setup, or low-signal wandering