podcast-analysis

name	podcast-analysis
description	Analyze podcast episodes from transcripts, audio URLs, or audio files. Use when the user mentions podcast, episode, audio transcript, interview, or provides a podcast file/URL.

Podcast Episode Analysis

Analyze podcast episodes to generate structured reports with summaries, guest insights, key quotes, and actionable takeaways. Supports transcript files, audio URLs (YouTube, podcast platforms), and audio files.

When to Use

Activate this skill when the user:

• Mentions "podcast", "episode", "audio", "interview"
• Provides a transcript file path (.txt, .srt, .md)
• Provides an audio file path (.mp3, .m4a, .wav)
• Provides a podcast URL (YouTube, Apple Podcasts, Spotify, etc.)
• Asks to summarize a podcast they listened to
• Wants to extract insights from an interview
• Mentions a specific podcast show or episode

Supported Inputs

• Transcript file: .txt, .srt, .md file with episode transcript
• Audio URL: Direct audio link or podcast platform URL (will be transcribed)
• Audio file: .mp3, .m4a, .wav file (will be transcribed)

Instructions

1. Determine input type:

   • If file path with text extension (.txt, .srt, .md): Read transcript directly
   • If audio file path (.mp3, .m4a, .wav, etc.): Transcribe with Whisper
   • If URL provided: Download and transcribe audio with Whisper
   • Ask user for input if not provided

2. For audio transcription (URLs or audio files):

   • Use yt-dlp to download audio (supports many podcast platforms)
   • Transcribe using OpenAI Whisper API (requires OPENAI_API_KEY in .env)
   • Fallback to local Whisper if available
   • Note: Transcription may take a few minutes for long episodes

3. For transcript files:

   • Verify file exists and is readable
   • Read the transcript content
   • Parse SRT format if applicable

4. If content retrieval fails:

   • Inform user about the issue
   • For audio: Check OPENAI_API_KEY is set
   • For files: Suggest checking file path
   • Stop here

5. Read the analysis prompt from prompts/podcast.md

6. Extract episode metadata:

   • Podcast name
   • Episode title
   • Host(s) and guest(s)
   • Duration if mentioned

7. Generate analysis following the prompt structure exactly

8. Create output directory reports/podcasts/ if needed

9. Save the report to reports/podcasts/YYYY-MM-DD_sanitized-title.md

10. Update the activity log at logs/YYYY-MM-DD.md:

    • Add entry under "## Podcasts Listened" section
    • Format: - [Title](../reports/podcasts/filename.md) - HH:MM

11. Confirm to user what was saved

Report Format

# [Episode Title]

**Podcast**: [Show Name]
**Host(s)**: [Names]
**Guest(s)**: [Names if any]
**Source**: [file path or URL]
**Date**: YYYY-MM-DD
**Type**: Podcast Episode

---

[Analysis content following prompts/podcast.md structure]

---

## My Notes

[Empty space for user notes]

Error Handling

• If file doesn't exist: Ask for correct path
• If transcript too short: Note but analyze
• If no speaker identification: Best effort analysis
• If prompts/podcast.md missing: Use prompts/default.md
• If audio download fails: Check yt-dlp is installed (pip install yt-dlp)
• If transcription fails: Check OPENAI_API_KEY is configured in .env
• If audio file too large: OpenAI API limit is 25MB, suggest splitting

Requirements for Audio

• yt-dlp for downloading audio (pip install yt-dlp)
• OpenAI API key for transcription (set OPENAI_API_KEY in web/backend/.env)
• Alternatively: Local Whisper installation (pip install openai-whisper)

Related

• Slash command: /podcast <filepath-or-url>
• Prompt file: prompts/podcast.md
• Output: reports/podcasts/
• Transcription service: web/backend/services/transcription.py