Your Podcast Transcript Is Worth More Than You Think
You recorded an episode. Descript transcribed it. Now what?
Most people rewrite the same content five different ways, or they don’t bother — and the episode sits on Spotify while the transcript collects dust in a folder. There’s a better way: one Claude command, and eight pieces of content come out the other side.
By the end of this post, you’ll have the exact skill file that takes a raw transcript and automatically generates show notes, Apple/Spotify descriptions, a Twitter thread, a LinkedIn post, a newsletter excerpt, a blog post, and a searchable knowledge base entry — all in one pass.
Why Transcript Repurposing Is a Pattern Problem
Show notes take 20 minutes. A blog post is an hour. A newsletter excerpt is another 20. A LinkedIn post on top of that. Then Twitter. Then the podcast platform descriptions.
It’s the same content, reformatted five times. That’s exactly the kind of repetitive, structured work agents were built to eliminate.
The transcript is the raw material. Every output format is just a different pattern applied to the same source. Claude is very good at pattern matching — so let it do the pattern work while you focus on the recording itself.
This is the same principle behind every Claude Code system I’ve built on this channel. If you’re new to running Claude Code inside VS Code, the Claude Cowork setup walkthrough covers the workspace structure you’ll need before following the steps below. And if you haven’t seen the research agent video, that’s a good companion — it handles the discovery side of content, while this skill handles the distribution side.
One note on transcription sources: this works with any transcript. Descript is the most common starting point because it transcribes automatically when you finish recording and exports with one click (Transcript → export as .txt). But Whisper, Otter, Rev, or any plain text transcript works equally well.
The 8 Outputs
Here’s what the agent generates from a single transcript run, in order:
- Show notes — 250–400 words, episode page format. Written to make someone who hasn’t listened want to press play — not a list of talking points.
- Apple Podcasts / Spotify descriptions — two versions: a 150-character search result hook and a 600-character episode description. Hard character limits enforced before saving.
- Chapter timestamps — two formats in one file: plain text for your show notes page, and YouTube chapter format if you cross-post the episode.
- Twitter / X thread — 6–8 numbered tweets. The opening tweet is the episode’s single most surprising or useful claim. Every tweet stands alone. Final tweet has the episode link, no hashtags.
- LinkedIn post — insight-led opener, different angle from the thread (not the same hook copy-pasted), three short paragraphs plus hashtags.
- Newsletter excerpt — 150–250 words that drop straight into the “what’s new this week” section of any newsletter. Starts with a hook, not “This week on the podcast…”.
- Blog post — a full standalone article, not a transcript dump. The content is restructured for a reader, not a viewer — 800–1200 words.
- Knowledge base entry — structured for Notion or Obsidian with title, summary, key insights, tools referenced, and quotable moments pulled from the transcript.
All eight save to a Podcast repurposer outputs/ folder as separate markdown files — all timestamped the same minute.
The Knowledge Base Angle
The output most podcasters don’t think about until episode 50 is the knowledge base entry.
Every episode automatically adds a structured, searchable record to an internal library. The entry captures the episode’s key insights (bolded), any frameworks or models mentioned, tools referenced, and exact quotable moments from the transcript. Tag vocabulary stays consistent across entries so the library is searchable by topic.
If you’ve been podcasting for two years, you have 50 or 100 episodes of IP locked in audio files. This builds a searchable archive of everything you’ve said — so you can find what you said about pricing in episode 34, or which three episodes touched on client onboarding, without rewinding audio.
It works in Notion, Obsidian, or any markdown-based tool.
How to Set This Up
Once the files are in place, the entire system runs with one phrase: “Repurpose this podcast episode” — paste your transcript, and Claude handles the rest.
Step 1: Install Claude Code
Open VS Code, search “Claude Code” in the Extensions marketplace, and install Anthropic’s official extension. Open a folder as your workspace — this is the directory Claude operates in.
Step 2: Create the folder structure
Inside your workspace, create:
- A
skills/folder - A file called
podcast-repurposer.mdinside it (paste the skill file below) - A folder called
Podcast repurposer outputs/
Step 3: Create a CLAUDE.md file
CLAUDE.md is the workspace context file — Claude reads it at the start of every session. Add this block to yours (create the file in the root of your workspace if it doesn’t exist):
## Podcast Repurposer
When asked to "Repurpose this podcast episode", run the skill in `skills/podcast-repurposer.md`.
Provide: transcript (paste or file path), episode title, episode number (or "N/A"), guest name and bio (or "Solo episode"), episode URL (or "[URL TBC]").
All outputs save to `Podcast repurposer outputs/`.
Step 4: Add your brand voice (recommended)
Create a file at brand/voice-guide.md and describe how you sound — your tone, what language you use, what you avoid. The skill reads this before writing anything. If you skip this step, Claude defaults to a clear, direct, conversational style.
Step 5: Run it
Open Claude Code, say “Repurpose this podcast episode”, and paste your transcript. Claude works through all 8 outputs in order, saves each file, and presents a summary when it’s done.
The Podcast Repurposer Skill File (Copy This Into Claude Code)
Save this as skills/podcast-repurposer.md in your workspace.
# Skill: Podcast Repurposer
> Takes a podcast (or any audio recording) transcript and generates all content outputs in one pass. Works with any transcript source: Descript, Whisper, Otter, Rev, or plain text.
## How to Invoke
Say: *"Repurpose this podcast episode"* and paste the transcript directly, or provide a file path.
**What you need to provide:**
1. The transcript — paste directly or point to a `.txt` or `.srt` file
2. Episode title
3. Episode number (or "N/A")
4. Guest name and one-line bio (or "Solo episode")
5. Episode URL (leave as `[URL TBC]` if not yet published)
---
## Context to Read First
Before generating anything, read:
- `/brand/voice-guide.md` — every output must match the host's tone (if this file doesn't exist, write in a clear, direct, conversational style)
- `/brand/audience.md` — write for your target audience (if this file doesn't exist, write for business owners and content creators)
---
## Outputs to Generate (in this order)
Work through each output one at a time. Save each to `Podcast repurposer outputs/` using the filename shown. After all are done, present a summary.
---
### 1. Show Notes
- Save to: `Podcast repurposer outputs/show-notes.md`
- 250–400 words, episode page format
- Lead with the most interesting premise, not "In this episode we discuss..."
- Key topics as 3–5 specific bullets — not vague summaries
---
### 2. Apple Podcasts / Spotify Descriptions
- Save to: `Podcast repurposer outputs/apple-spotify-descriptions.md`
- Two versions:
- **Short (150 chars max):** hook for search results — count characters before saving
- **Full (600 chars max):** episode page description, plain text only (no markdown)
- Both versions: lead with the insight, not the format
---
### 3. Chapter Timestamps
- Save to: `Podcast repurposer outputs/chapter-timestamps.md`
- Read the transcript and identify natural topic transitions
- Produce two formats in the same file:
- **Plain text** (for show notes): `00:00 Introduction`
- **YouTube chapters** (if cross-posting): `0:00 Introduction`
- Estimate timestamps proportionally if exact times are not available
- Note at the top of the file if timestamps are estimated
---
### 4. Twitter / X Thread
- Save to: `Podcast repurposer outputs/twitter-thread.md`
- 6–8 tweets, numbered (1/7, 2/7 etc.)
- Tweet 1: the most surprising or useful single claim from the episode
- Tweets 2–(N-1): one insight each, each standing alone
- Final tweet: one link (episode URL) — no hashtags
- Every tweet must be under 280 characters — count before saving
---
### 5. LinkedIn Post
- Save to: `Podcast repurposer outputs/linkedin-post.md`
- Insight-led opener — no "I" as the first word
- Different angle from the Twitter thread — don't repeat the same hook
- 3 short paragraphs + link + 2–3 hashtags at the bottom
---
### 6. Newsletter Excerpt
- Save to: `Podcast repurposer outputs/newsletter-excerpt.md`
- 150–250 words
- Opening hook: surprising claim or question — not "This week on the podcast..."
- Body: 2–3 sentences, one concrete detail
- CTA: one link, direct
---
### 7. Blog Post
- Save to: `Podcast repurposer outputs/blog-post.md`
- Full standalone article — not a transcript dump, not a summary
- Restructure content into logical written sections with clear headings
- 800–1200 words
- Key rule: spoken language ≠ written language — adapt, don't transcribe
---
### 8. Knowledge Base Entry
- Save to: `Podcast repurposer outputs/knowledge-base-entry.md`
- Structure: title, summary, key insights (bolded), frameworks mentioned, tools referenced, quotable moments (exact transcript quotes only), related episodes
- Use consistent topic tags — check existing entries before creating new ones
---
## Important Rules
- **Each output is platform-specific** — LinkedIn and the Twitter thread should have different hooks and angles. Don't copy the same content across platforms.
- **Adapt, don't transcribe** — especially for the blog post and show notes. Spoken language ≠ written language.
- **Be specific** — pull exact details, numbers, and claims from the transcript. Generic outputs are useless.
- **Character limits are hard limits** — Apple Podcasts at 151 chars is rejected. Twitter at 281 chars gets cut. Count before saving.
- **Save to `Podcast repurposer outputs/`** — one file per output.
---
## After Generating All Outputs
Present a summary:
\`\`\`
## Content Package: [EPISODE TITLE]
✅ Show Notes — saved
✅ Apple/Spotify Descriptions — saved (150 chars + 600 chars)
✅ Chapter Timestamps — saved
✅ Twitter Thread — saved (N tweets)
✅ LinkedIn Post — saved
✅ Newsletter Excerpt — saved
✅ Blog Post — saved
✅ Knowledge Base Entry — saved
📁 All files in: Podcast repurposer outputs/
📌 Review notes:
- [Any flags: estimated timestamps, missing guest info, URLs still TBC, etc.]
\`\`\`
What Comes Next
This setup gives you the content. Getting it out the door is a separate step — and one I’ve already built.
The Blotato scheduling pipeline takes content files like these and automatically schedules them to LinkedIn, Twitter, Instagram, TikTok, and YouTube — without touching a browser. Claude reads the transcript via Blotato, writes platform-specific captions, and schedules each post using the Blotato API, all from a single conversation.
Put the podcast repurposer and the Blotato pipeline together and you go from transcript to published content across every platform with no manual steps at all. Record once. Run the agent. Done.
Final Thoughts
One transcript. One command. Eight outputs — show notes, platform descriptions, a Twitter thread, a LinkedIn post, a newsletter excerpt, a blog post, and a searchable knowledge base entry.
The skill file above is everything you need to get started. Copy it into your workspace, follow the five setup steps, and run it on your next episode.
If you want to see it running live — and understand how the knowledge base output works in practice — watch the full video walkthrough here.