🐛 Spoken
Search for podcast episodes and fetch their transcripts as clean markdown with speaker names and timestamps. Uses the spoken.md API.
About Spoken
Search for podcast episodes and fetch transcripts as clean markdown with speaker names and timestamps.
This medium 329-word instruction set is purpose-built for debugging work in AI coding agents. Paste directly into your CLAUDE.md or AI agent settings.
Key Capabilities
- If the user provided a numeric episode ID, skip to step 2
- If the user provided a URL (Spotify, YouTube, etc.) or a text description, search for it:
- `GET https://spoken.md/search?q={query or URL}`
- Include header `x-api-key: {SPOKEN_API_KEY}` (use `pt_demo` if no key is set - returns demo episode only)
- Show the user the results and let them pick, or pick the best match if unambiguous
What This Skill Covers
- Steps
- Transcribing a whole show
How It Works
- Find the episode:
- Fetch the transcript using the WebFetch tool:
- Handle the response:
- Get the show's `podcastId`. Every `/search` result includes one.
Use Cases
- Systematically diagnosing production errors and crashes
- Tracing memory leaks and performance bottlenecks
- Debugging async race conditions and concurrency issues
- Root-cause analysis with structured logging
Example Prompts
System Prompt (329 words)
Podcast Transcript (spoken.md)
Search for podcast episodes and fetch transcripts as clean markdown with speaker names and timestamps.
Steps
- Find the episode:
GET https://spoken.md/search?q={query or URL}
- Include header x-api-key: {SPOKEN_API_KEY} (use pt_demo if no key is set - returns demo episode only)
- Show the user the results and let them pick, or pick the best match if unambiguous
- If unclear what episode they want, ask- Fetch the transcript using the WebFetch tool:
https://spoken.md/transcripts/{id}
- Include header x-api-key: {SPOKEN_API_KEY} (use pt_demo if no key is set - demo episode only)- Handle the response:
top_up_url - they need to POST to it to buy more credits
- 404: Tell the user no transcript was found for that episode
- 502: Upstream error - suggest retrying in a moment- If the response includes
X-Credits-Remainingheader, mention how many credits remain. Note:X-Credits-Charged: 0means this episode was already fetched before (repeat fetches are free).
Transcribing a whole show
To transcribe a podcast's entire back-catalog (e.g. "get me every episode of X"):
- Get the show's
podcastId. Every/searchresult includes one. - List every episode:
GET https://spoken.md/podcasts/{podcastId}/episodeswith thex-api-keyheader. This is free (metadata only) and returns a JSONepisodesarray, each with anid,title, anddate. - Fetch each
idthroughhttps://spoken.md/transcripts/{id}as in step 2 above. Each new episode costs 1 credit; repeat fetches are free. Make sure the key has enough credits for the whole show before starting.
Frequently Asked Questions
What is Spoken?
Spoken is a free debugging skill for AI coding agents. Search for podcast episodes and fetch their transcripts as clean markdown with speaker names and timestamps. Uses the spoken.md API.. It provides a specialized system prompt that configures your agent with debugging expertise.
How do I use Spoken with Claude Code?
Copy the system prompt below and paste it into your CLAUDE.md file or Claude Code's custom instructions to activate Spoken.
Which AI coding agents work with Spoken?
Spoken is compatible with Claude Code, Cursor, GitHub Copilot, Windsurf, OpenClaw, Cline, and any AI agent that supports custom system prompts or .cursorrules files.
Is Spoken free to use?
Yes, Spoken is completely free and open source. The full source is available on GitHub at https://github.com/spokenmd/spoken/blob/master/SKILL.md. You only need a subscription to the AI agent you use it with.
